Beyond AI Citations: How AI Search Decides Which Brands to Recommend

SUPERCHARGE YOUR ONLINE VISIBILITY! CONTACT US AND LET’S ACHIEVE EXCELLENCE TOGETHER!

A brand can appear in an AI-generated answer, earn a source citation and still lose the customer. That sounds contradictory if your reporting dashboard is built around Google rankings, organic sessions and backlinks. It makes much more sense once you separate evidence used in an answer from the conclusion that answer delivers.

Imagine a procurement director asking an AI assistant for the best enterprise analytics software for a regulated business. Your company publishes a well-researched article on the subject. The assistant uses that article to explain a compliance requirement, then recommends three competitors. Your page helped construct the answer, but your business did not make the shortlist. Was the citation valuable? Potentially. Was it the commercial outcome you wanted? Not by itself.

Sources: Alphabet/Google Q2 2025 reporting; Previsible/Semrush summary; public AI search adoption reporting.

At ThatWare, we believe this distinction should change how brands plan SEO, generative engine optimization (GEO), answer engine optimization (AEO) and AI visibility measurement. We are not suggesting that rankings or citations have stopped mattering. We are saying that they describe only parts of a larger decision journey: discovery, retrieval, interpretation, comparison, attribution and recommendation.

A shorter version of this framework was presented by Dr. Tuhin Banik during our brightonSEO keynote. This expanded article explores the business problem, the research behind it, the practical architecture we recommend, and the measurements marketing teams can use when the answer is becoming as important as the click.

The central distinction: a citation is evidence, not endorsement

A citation typically shows that a system has associated a source with some part of its response. In a Google AI Overview, it may mean the content helped support a factual explanation. In a conversational assistant with web access, it may indicate a source consulted during that response. Neither behavior automatically proves that the system approves of every statement on the page or considers the company behind the page the best choice for the user.

That gap matters because three different questions are often confused in performance reports. Was our content found? Did the system use or cite our content? Did the system recommend our brand for the buyer’s need? The first is about discoverability, the second about answer influence and the third about selection. They may correlate, but they are not interchangeable.

Consider a commercial roofing manufacturer. An AI answer may cite its specification document for a fire-rating definition, yet recommend a different manufacturer for a coastal installation because the first brand’s website does not clearly establish coastal suitability, local availability or the relevant warranty terms. The citation is legitimate; the absent recommendation is also understandable. Fixing only the cited page’s wording would miss the missing commercial proof.

A stronger AI visibility strategy therefore asks what the system can verify about the organization as well as the specific document. When a buyer requests a shortlist, the answer must connect the brand to the right category, product capability, geography, evidence and buying context. A source document alone rarely carries all of those signals.

Figure 1. A citation is an output signal. It does not establish recommendation, trust across all claims, or repeatability across prompts. Credit: ThatWare brightonSEO presentation.

Why this matters now: AI answers have become a buying-stage discovery layer

The adoption curve is no longer hypothetical. Alphabet reported that Google AI Overviews had surpassed 2 billion monthly users in July 2025. A June 2026 Alphabet investor presentation put that number above 2.5 billion monthly users. These are Google’s reported usage figures, not independent estimates of unique people using every AI search product. Even with that qualification, the scale makes AI-mediated discovery a mainstream marketing consideration.

Yet visibility inside those answers does not behave exactly like ordinary website traffic. Pew Research Center analyzed browsing behavior among 900 US adults in March 2025. In the sampled Google visits, users clicked a traditional result link after seeing an AI summary in 8% of visits, versus 15% when no AI summary appeared. Clicking a source inside the AI summary occurred in only about 1% of visits with a summary. These figures describe Pew’s observed sample, not a universal forecast for every industry, country or query.

The signal for brands is straightforward: a useful answer can change a person’s understanding without producing a website session. For publishers, that may be a difficult trade-off. For a provider selling high-value services or products, it also means that shortlist inclusion, positive description, accurate comparison and branded follow-up searches deserve attention alongside clicks.

News provides an especially visible example. Similarweb’s research reported that the share of news-related Google searches ending without a click rose from approximately 56% in May 2024 to 69% in May 2025. This is category-specific US news-search evidence, not proof that every commercial query behaves in the same way. The practical lesson is that the traditional path from result impression to page visit is no longer the only path through which information influences a choice.

For a growth team, the wrong reaction is to abandon organic SEO. The better reaction is to extend measurement. We still want technically sound pages, qualified traffic and conversions. We also want to know whether relevant AI systems describe the business correctly and include it in the right commercial conversations.

Figure 2. The click-through gap presented in the keynote is based on Pew Research Center’s observed Google searches. Read sample-specific figures rather than treating them as platform-wide conversion rates. Graphic: ThatWare; data source: Pew Research Center.

From ranking to retrieval to selection: an expanded search model

For years, a simplified search program looked like this: win the query, earn the click, persuade the visitor on the landing page. That model remains important. It is simply incomplete when a user asks an AI assistant to recommend, compare or explain options before visiting a site.

An expanded model has three commercial milestones. Be discoverable: make authoritative information accessible through ordinary search infrastructure and wherever a specific AI product lawfully retrieves content. Be usable: present verifiable evidence that an answer can summarize accurately. Be suitable: make it clear why the brand is or is not right for the particular buyer. The last milestone often determines whether a business appears in a final recommendation rather than only in the sources.

A single generic service page is usually weak at the third milestone. Take the prompt, “Which enterprise GEO agency can help us audit AI citations across international markets?” A response would need more than a declaration of SEO expertise. It could reasonably seek relevant service detail, geographic capabilities, a documented audit methodology, reporting examples, knowledge of multiple AI products and an evidence trail for the claims.

This is why AI search optimization services should be evaluated as an integrated program rather than a batch of content rewrites. Technical SEO makes the material accessible. Semantic structure clarifies what the brand does. Expert content supplies useful answers. Third-party references provide external context. Commercial pages explain fit. Measurement reveals whether those investments change observable responses.

Figure 3. The new objective is not to replace traditional rankings but to add retrieval, answer presence and commercial selection to the measurement model. Credit: ThatWare brightonSEO presentation.

How an AI system can arrive at a brand conclusion

There is no single published algorithm that describes every model, every search engine and every response mode. Products differ in their retrieval indexes, browsing behavior, ranking systems, source controls, citation rules and personalization. Our six-stage diagram is therefore a diagnostic model, not a claim that we know the internal sequence of every AI provider.

Stage 1: discover and access. A public service page, product manual, review, news article, profile or structured data entry may become accessible to a search or retrieval system. Eligibility still depends on the provider and its technical requirements. If the content is blocked, inaccessible, thin, duplicated or available only in unusable interface elements, discoverability is harder.

Stage 2: retrieve. Given a question, the system may retrieve a set of candidate pages or passages that seem relevant. The candidate set is not necessarily identical to the top ten traditional search results. Which sources become candidates can change with query wording, context and freshness.

Stage 3: allocate attention. Among retrieved items, a system must decide what evidence is relevant enough to use. A comparison table containing exact product limitations may be more useful for one task than a thousand-word thought-leadership article, even if the article ranks strongly for broad keywords.

Stage 4: synthesize. The answer combines factual claims and context. At this stage, a brand’s product capabilities, geographic scope and third-party validation may become part of the generated explanation, or may be lost if supporting evidence is unclear.

Stage 5: attribute. Some systems attach source links to parts of the answer. Citation behavior varies and may be incomplete. A citation is best interpreted as an observable attribute of the answer, not as a universal quality grade for the cited business.

Stage 6: conclude. When the user asks for advice or a shortlist, the system may compare candidates and select a recommended option. This step may consider fit, constraints, trust and the information available in that interaction. Crucially, the brand that contributed evidence to the explanation need not be the brand selected for the purchase.

The business question becomes: at which stage is our brand being lost? A GEO audit that measures only citations may never discover that the real obstacle is an unclear service category, an unsupported claim or a missing comparison page.

Figure 4. A six-stage conceptual model for diagnosing brand visibility gaps: access, retrieval, context, synthesis, attribution and conclusion. Credit: ThatWare brightonSEO presentation.

A useful framework for thinking about brand selection

In the keynote, we illustrated brand conclusions through six evidence classes: entity clarity, evidence density, source consensus, freshness, experience signals and risk checks. These are practical planning categories, not a literal mathematical formula taken from a model provider. They help editorial, technical and commercial teams ask better questions about what a buyer can confirm.

1. Entity clarity: who are you, precisely?

If a business operates under several names, mixes product and service labels, or has outdated biographies across trusted profiles, an AI answer may connect facts to the wrong organization. Entity clarity begins with canonical brand naming, defined offers, leadership information, location and ownership relationships, consistent contact details, and clean navigation from topic pages to the right commercial page.

Clarity is not merely adding Organization schema. It is making the website and external references agree on what the company actually does. For a manufacturer, this may include the parent company, product families, certifications, use cases and dealer relationships. For an agency, it means distinguishing consulting, managed services, proprietary methodologies and training rather than using one vague “AI solutions” label everywhere.

2. Evidence density: can a reader verify the claim?

A sentence saying “we deliver outstanding results” adds little. An explained methodology, product specification, named certification, transparent service scope, dated example or independently verifiable result provides more substance. Not every page needs a research study. Every commercial claim should, however, be appropriate to the evidence available.

For instance, a product page stating that a tool has SSO, region-specific data residency and a defined retention policy offers facts a procurement team can compare. A similar page that says only “enterprise-grade security” leaves too much to inference. High-value, specific proof helps humans assess the offer and gives answer systems clearer material to work with.

3. Source consensus: what do independent sources say?

The brand’s website is an important primary source about its own services. But independent product reviews, industry directories, standards bodies, knowledgeable publishers, customer interviews and credible professional communities can clarify reputation or context. Different source classes deserve different weight. A verified certification is not equivalent to an anonymous forum comment, and a sponsored placement is not the same as a peer-reviewed study.

Our goal is not artificial mention volume. It is a consistent, factual external footprint. Purchasing inauthentic endorsements or manufacturing citations is unlikely to build durable buyer trust and may conflict with platform policies.

4. Freshness: is the information still true?

An assistant can misstate availability if it finds an old pricing page, discontinued product, expired qualification or outdated location listing. A disciplined freshness program updates important facts where users and systems encounter them: canonical service pages, documentation, downloadable materials, business profiles and relevant third-party listings.

Not all content needs a new date every month. Evergreen definitions may change slowly. Pricing, service geography, certifications, inventory and product integrations may need much tighter review cycles. Prioritize refreshes according to the cost of getting the answer wrong.

5. Experience: can you demonstrate delivery?

Practical experience is often visible through worked examples, accountable expert authorship, methodology documents, screenshots of genuine features, process explanations, responsibly anonymized cases, client-approved testimonials and specific implementation lessons. The most persuasive information is rarely “we are experts.” It is evidence that a prospective buyer could use in an evaluation.

An AI answer about the best vendor for a complicated requirement benefits from these concrete distinctions. So does the person reading the underlying page. Invest in material that survives both a human procurement review and a machine-generated summary.

6. Risk and suitability: should this brand be recommended for this need?

A brand may have excellent content and still be inappropriate for a particular buyer. Geography, regulatory restrictions, budget, product limits, safety concerns, minimum contract size and service exclusions are legitimate reasons for a recommendation to change. A responsible visibility program does not try to suppress those constraints. It explains them accurately.

That is why AI brand reputation management must be tied to truthful product information. The ideal result is not a positive recommendation in every prompt. It is a correct, well-supported recommendation when the organization is a strong fit, and an accurate explanation when it is not.

Figure 5. ThatWare’s six evidence classes are an audit heuristic for brand understanding, not a disclosed scoring formula used by any AI provider. Credit: ThatWare brightonSEO presentation.

Five evidence environments that shape an AI-facing brand footprint

When evaluating LLM visibility services, it helps to look beyond the corporate website. A potential answer can draw context from several evidence environments, depending on the platform, the query and the material available.

Owned evidence includes product and service pages, technical specifications, original research, documented methodologies, FAQs, data sheets, policy pages and case studies. This is where a business controls its clearest statement of fact. The priority is accuracy, accessibility and internal consistency.

Earned evidence comes from independent reporting, credible citations, relevant awards, third-party evaluations, professional publications and genuine analyst commentary. It can place the brand within a wider market, but brands should avoid treating any press mention as equivalent to a trusted review of their offering.

Community evidence appears in reviews, practitioner discussions, public Q&A, video demonstrations, forum threads and social conversations. These materials may reveal customer language and recurrent objections. Their quality is uneven, which is why a marketer should investigate claims rather than automatically amplify them.

Structured evidence includes accurate schema markup, canonical entity relationships, product feeds where applicable, stable author pages, verified profiles and consistent organizational metadata. Structure assists interpretation and certain search features, but it is not a separate “AI recommendation switch.”

Behavioral and commercial context includes the questions buyers ask, how they search for brands, the terms used to compare products, and conversion signals a company can observe in its own analytics. These do not prove which private model inputs caused a particular answer. They help marketers choose the prompts and content priorities that actually match customer demand.

Think about these as an evidence portfolio. The right strategy is not to maximize every bucket equally. A B2B SaaS company with missing integration documentation needs different work from a local specialist whose problem is inconsistent location information. A meaningful AI citation audit starts with the specific buyer questions the brand wants to answer.

Figure 6. Five evidence environments that can inform a brand’s public footprint: owned, earned, community, structured and behavioral. Credit: ThatWare brightonSEO presentation.

Mentioned, cited, described, compared and recommended are different outcomes

A dashboard that records “brand appeared: yes” makes little distinction between a passing mention and a strong commercial recommendation. We recommend recording the outcome more precisely. Was your brand cited as an evidentiary source? Named in a list? Described accurately? Compared against alternatives? Recommended for the requested use case? The distinctions tell a very different performance story.

Here is a practical example using a fictional warehouse automation provider. In one prompt, the assistant cites a company whitepaper about RFID deployment. In another, it lists the business among providers without explaining why. In a third, it describes its hardware and integration scope correctly. In a fourth, it compares that provider with two competitors on deployment time and compatibility. In a fifth, it places the business on the shortlist for a defined warehouse size and budget.

All five answers contain visibility. Only some contain commercially useful selection. The quality of the answer also matters: a recommendation based on false product claims is not a success simply because it sounds positive.

For internal reporting, avoid collapsing these into one overconfident “AI rank.” Track separately the presence rate, source citation rate, factual accuracy, comparison inclusion, qualified recommendation rate and negative or inappropriate recommendation rate. Report how often results change when the same intent is tested with different prompts and at different times.

We often phrase the target as answer presence with a supported, positive conclusion. The word supported is important. Recommendation quality is stronger than mention volume, and honesty is stronger than a flattering hallucination.

Figure 7. The presentation’s outcome ladder separates citation, mention, description, comparison and recommendation. Credit: ThatWare brightonSEO presentation.

What independent research tells us about citation selection

Several research findings help explain why an SEO position cannot simply be converted into an AI citation forecast. The evidence is promising, but the limits are just as important as the headline numbers.

A 2026 study by Haofei Xu, Umar Iqbal and Jacob M. Montgomery, Measuring Google AI Overviews: Activation, Source Quality, Claim Fidelity, and Publisher Impact, examined 55,393 trending queries across 19 categories over a defined 40-day period. In that sampled corpus, AI Overviews appeared for 13.7% of queries overall and 64.7% of question-form queries. Nearly 30% of cited domains did not appear in the simultaneously displayed first-page organic results. These are findings from a particular query set, not universal activation rates for all searches.

The authors also assessed 98,020 atomic claims and found 11% were unsupported by the cited pages under their methodology. This matters for brand governance. A citation beside a statement is not automatic proof that the linked page validates that specific claim. A business should inspect the exact wording of important AI responses, especially when product compatibility, safety, pricing or compliance are involved.

The study strengthens a point we made in the keynote: retrieval, ranking, citation and fidelity are separate diagnostic questions. A first-page ranking may help a brand, but there is no defensible rule that says the number-one result must be cited, or that a cited source must be the eventual recommendation. The research also gives marketers a reason to track source-to-claim support rather than counting links as interchangeable units.

Another relevant paper, *Generative Engine Optimization* by Aggarwal and colleagues, reported visibility gains of up to 40% in its benchmark setting from tested content strategies. That figure is an experimental result under specific conditions, not a promised 40% uplift for a real business. Academic research helps identify plausible interventions; only disciplined testing can show whether they work for a particular website, market and platform.

A 2026 critical survey of GEO research also highlights variability, weak transferability across settings and the danger of treating citation improvement as the same thing as durable discoverability or business impact. We regard that caution as useful, not inconvenient. It protects a marketing budget from oversimplified promises.

Figure 8. The 2026 AI Overviews study shows why first-page ranking, citation selection and claim support must be assessed separately. Graphic: ThatWare; underlying research: Xu, Iqbal and Montgomery, arXiv 2026.

Nine practical changes that make a brand easier to evaluate

Most brands do not need a dramatic website rebuild to begin improving AI-facing evidence. They need a prioritized set of fixes based on gaps that affect both machine-readable clarity and real buyer confidence. Here is how we would approach the work in an AI search visibility audit.

1. Give every important service or product an unmistakable identity

Build a register of the services, products, industries, locations and named methodologies that matter commercially. For each item, identify the canonical page, accepted name, short definition, parent category, unique attributes, relevant evidence and conversion action. Remove contradictions across the homepage, navigation, brochures and third-party profiles.

If a company offers both consulting and implementation, explain that distinction. If a product is available only through distributors, say so. If a service has three labels in different departments, choose the public canonical wording and connect legitimate synonyms without duplicating pages. This is where entity SEO services become operational rather than decorative.

2. Write claims that survive being separated from their page

An answer system may use a passage without carrying the entire website’s context into the response. Make important claims explicit, qualified and easy to verify. Instead of “our solution handles every integration,” name the supported systems, the limitations and the documentation. Instead of “the most secure platform,” describe the relevant controls and verifiable certifications.

This is good copywriting for procurement committees as well as AI responses. It reduces ambiguity when a paragraph is quoted, summarized or compared. Keep crucial qualifying information near the claim rather than hiding it in a disconnected legal page.

3. Build comparison content for real buying decisions

Commercial intent often arrives as a comparison: an agency versus an in-house team, a hosted platform versus open source, a local provider versus a national chain, or one product tier versus another. Many companies write hundreds of informational blogs while leaving their buying questions unanswered.

Create useful comparison pages that disclose criteria, trade-offs, ideal-fit conditions and legitimate limitations. Avoid attacking named competitors with unsupported assertions. A fair “who this is for” section can be more persuasive than a table engineered to make your own offering win every row.

4. Show original evidence rather than generic reassurance

Publish authentic process examples, product demonstrations, original research methods, annotated screenshots, customer-approved outcomes and detailed implementation guides. Explain what was measured, what changed, what was not measured and what could limit generalization. A small but transparent example often carries more credibility than a sweeping percentage without its denominator.

For regulated sectors, obtain appropriate permissions and protect personal or confidential information. The goal is verifiable expertise, not the appearance of a case study.

5. Repair internal information architecture

A brand’s important proof may exist but be difficult to discover. Product specifications live in PDFs; FAQs are disconnected from service pages; author biographies are buried; regional variants conflict. Develop internal links that connect a buyer’s problem to the right explanation, proof and action. Give each commercial page a clear place within its topic cluster.

Search engines and human readers benefit from predictable hierarchies. We typically map informational content to decision pages with descriptive links. Do not force an irrelevant sales CTA into every paragraph. Make the next logical step visible when the reader is ready.

6. Make technical access boringly reliable

Confirm that important URLs can be crawled and indexed where appropriate, that important text is available in usable HTML, that canonical tags are correct, and that JavaScript does not hide the content required to understand the offer. Test structured data against visible page content. Review robots rules, noindex directives, internal link accessibility, page performance and PDF alternatives.

For Google AI Overviews and AI Mode, Google Search Central’s guidance says established SEO fundamentals remain relevant. There is no proprietary AI-specific schema that a site must add to become eligible. A good technical foundation is more defensible than a list of speculative tags.

7. Align the external brand footprint

Check high-value directories, professional profiles, supplier listings, relevant publications, product marketplaces and public customer resources. Are they describing the same services, locations, names and capabilities? Are outdated credentials still being repeated? Are independent references accurately linked to the relevant offering rather than just the homepage?

Prioritize corrections that affect customer trust and category understanding. Do not turn this into a campaign for indiscriminate mentions. Quality, topical fit and factual consistency are more useful than sheer volume.

8. Create an explicit content-maintenance process

Assign owners to volatile facts: price ranges, service boundaries, compliance standards, version numbers, warranties, opening hours and supported integrations. Record where those claims appear in HTML, downloads, schema and external profiles. When a business changes an offering, update the canonical evidence before publishing another promotional article that repeats the old message.

This is the operational side of AI brand trust. A brand cannot expect consistent public descriptions while publishing contradictory versions of its own story.

9. Measure outcomes against a controlled prompt set

Pick a manageable set of commercial, comparison, informational, local and brand-led questions, then test them repeatedly. Save the exact prompt, date, location setting when relevant, platform, answer text, cited URLs, brand position, accuracy issues and recommendation result. Include the main competitors so the evidence explains relative performance.

Testing should not be a one-off screenshot exercise. A single answer can change with phrasing, account context or retrieval state. Sample repeatedly, report uncertainty and assess whether improvements are visible across meaningful categories rather than one flattering prompt.

Build a brand knowledge layer before chasing AI tricks

A well-maintained brand knowledge layer is a practical collection of canonical facts and evidence that helps customers, employees, partners and retrieval systems understand what a company offers. It is not a secret file that guarantees model training or ranking. Its value comes from reducing ambiguity in public information and making that information easier to maintain.

We organize the work into five areas. First is the entity spine: the approved organization name, product and service inventory, leadership information, locations, aliases and relationship definitions. Second is answer-ready material: useful explanations of real customer questions with nearby evidence and clear qualifications. Third is the evidence graph: a map from each significant claim to its supporting documents and independent references. Fourth is accessible structure: a logical hierarchy, internal links, properly implemented metadata and formats that real users can read. Fifth is a freshness loop: named owners, revision schedules and a record of what changed.

For a B2B technology company, this may be a collection of product pages, supported integration lists, security documentation, implementation guides, examples, pricing explanations and an accurate company profile. For a local professional service, it may be credentials, locations, availability, service boundaries, transparent process information and carefully maintained directories. Different industries need different evidence; the architectural principle stays the same.

A valuable side effect is internal alignment. Marketing, sales, customer success and product teams stop making different versions of the same claim. Customers receive better answers even before an AI assistant is involved. That is a sound return on an information-governance investment regardless of future platform changes.

Figure 9. The brand knowledge layer joins canonical identity, relevant answers, verifiable evidence, accessible structure and content freshness. Credit: ThatWare brightonSEO presentation.

What about llms.txt, AI manifests and custom knowledge files?

The presentation also discussed optional machine-readable assets such as llms.txt, internal entity inventories, research indexes, AI manifests, knowledge graphs and documented feeds. These can be useful in proprietary workflows or for agents that explicitly support them. However, brands should not confuse a proposed convention or an internally defined filename with a search-engine requirement.

Google’s official generative AI search guidance explicitly says websites do not need to create extra AI text files or special markup to appear in its AI features. The same guidance warns against treating tactics such as llms.txt creation, artificial mentions or content restructuring gimmicks as substitutes for fundamentals. That is an important boundary for any agency promising “AI-ready” file installations.

Our position is pragmatic. A structured entity register, evidence index or feed can be useful because teams and supported software can maintain and consume it. The file should have a documented consumer, schema, owner and validation process. If no system reads a proposed file, the expected benefit should not be marketed as proven. Always prioritize accurate public HTML, crawlable core pages, ordinary structured data where relevant, reputable evidence and robust website architecture.

Some future-facing capabilities will evolve. Open standards may mature and particular agents may adopt specialized protocols. We are interested in testing those developments, but testing is not the same as claiming every AI platform honors the same collection of files. The durable investment is the quality and accessibility of the facts those files contain.

Figure 10. Selected experimental machine-readable assets from ThatWare’s foundation-layer illustration. These are proposed or workflow-specific assets, not mandatory Google AI search files. Credit: ThatWare presentation.

SEO, AEO and GEO work best as one commercial program

These terms describe overlapping disciplines, not sealed silos. SEO helps people and search systems discover reliable pages. AEO focuses on making useful answers clear and accessible for direct-answer experiences. GEO focuses on how content and credible brand evidence may appear within AI-generated responses. A commercially useful program combines all three while keeping claims about influence proportionate to observable evidence.

An enterprise software business might begin with conventional technical SEO because its documentation is not indexed correctly. A healthcare provider may need carefully written, qualified answers to patient questions, along with accurate clinician and location information. A multinational manufacturer may need better entity mapping across product lines, technical documents and distributor markets. In each case, the mix of SEO, AEO and GEO should be determined by the bottleneck rather than a fixed package of fashionable deliverables.

ThatWare’s GEO services, AEO services and LLM SEO services approach these problems through complementary activities such as technical accessibility, semantic clarity, evidence-led content and visibility testing. A credible GEO agency should be able to explain the difference between a page that ranks, a source that gets cited and a brand that earns a qualified recommendation.

In particular, do not trade reader usefulness for robotic prose. Google’s published advice emphasizes helpful, distinctive content and the continued value of core SEO. The practical target is useful, well-evidenced information, not text written to match an imagined token pattern in a model.

Figure 11. SEO, AEO and GEO have distinct emphases, but the strongest implementation integrates discoverability, direct answers and sourced brand understanding. Credit: ThatWare brightonSEO presentation.

Our proven success record in SEO, AEO, GEO and LLM SEO

SEO

How Antonio’s Towing Transformed Their SEO and Digital Presence: A Deep Dive

In today’s highly competitive digital landscape, having a strong online presence is critical for service-based businesses, especially in niche industries like towing and equipment transport. Antonio’s Towing, a company specializing in towing, recovery, and low boy equipment transport, faced a common challenge. Despite offering high-quality services, their website struggled to attract organic traffic, and most leads came from paid search campaigns. This reliance on advertising resulted in higher costs and limited long-term growth potential.

To address this challenge, Antonio’s Towing partnered with ThatWare to implement a comprehensive SEO and digital transformation strategy. The main objectives were to improve search engine visibility, increase organic traffic, and drive meaningful conversions while ensuring the website delivered a seamless user experience.

This case study explains how a data-driven, holistic approach to SEO including search intent optimization, technical SEO improvements, content updates, and backlink strategies significantly enhanced Antonio’s Towing’s online presence. It also highlights measurable results, key lessons, and actionable strategies that can help other service businesses improve their digital footprint and achieve sustainable growth in competitive markets.

Understanding the Problem: The Client’s Initial Situation

Low Organic Visibility

Antonio’s Towing faced a significant challenge with its online presence. Despite offering high-quality towing and equipment transport services, the website was not attracting substantial organic traffic. Search engines were not effectively recognizing the site, which meant that potential customers searching for services like heavy equipment transport or towing in Michigan often did not encounter the company’s website in search results. This low visibility limited opportunities for lead generation and reduced the brand’s overall reach in the digital market.

Heavy Dependence on Paid Search

Because organic traffic was minimal, the business relied heavily on paid search campaigns to generate leads. While pay-per-click advertising delivered immediate visibility, it also meant that growth was tied directly to marketing spend. Any reduction in advertising budgets could lead to a sharp decline in website traffic and conversions. Additionally, paid search does not provide long-term, sustainable results, making it difficult for the business to build a strong online presence or establish authority in its niche.

Challenges of a New Website in a Competitive Market

The website was relatively new, with a domain age of less than a year. New domains often face challenges in gaining trust and authority with search engines. With a Domain Authority of just 4 and a limited backlink profile, the site struggled to compete with established competitors. In addition, the towing and heavy equipment transport market is highly competitive, with multiple regional and national players targeting the same high-intent keywords. These factors made it difficult for Antonio’s Towing to achieve meaningful organic visibility without a strategic, targeted approach.

Technical and Content Limitations

Beyond external competition, the website also had internal limitations that impacted performance. Technical SEO elements such as page speed, site structure, and crawlability required optimization to meet modern search engine standards. Additionally, the site content was not fully aligned with search intent. While it described services offered, it did not fully address the specific questions and concerns potential customers might search for online. This gap meant missed opportunities to capture high-intent traffic and improve engagement metrics.

The Need for a Comprehensive Solution

These challenges highlighted the importance of a holistic SEO and digital strategy. Antonio’s Towing needed a solution that combined technical improvements, content optimization, and a long-term approach to off-page SEO. By addressing both the technical and strategic aspects of digital marketing, the business could reduce reliance on paid campaigns, improve organic visibility, and establish itself as a trusted authority in its niche.

Technical SEO & User Experience (UX) Enhancements

The Role of Technical SEO in Modern Websites

Technical SEO serves as the foundation of a website’s search engine performance. Even the most compelling content may fail to rank if technical elements prevent search engines from efficiently crawling and indexing pages. For Antonio’s Towing, the site had several technical limitations that impacted both visibility and user experience. ThatWare identified key areas for improvement, including page speed, site structure, mobile usability, and Core Web Vitals. Addressing these factors was essential for creating a website that performed well in search engines and delivered a seamless experience to users.

Improving Core Web Vitals

Core Web Vitals are a set of metrics that measure user experience and page performance. They include:

• First Contentful Paint (FCP): Time until the first visible content appears.
• Largest Contentful Paint (LCP): Time until the largest content element loads.
• Total Blocking Time (TBT): Time during which the page is unresponsive to user input.
• Cumulative Layout Shift (CLS): Measures visual stability during page load.

In Antonio’s Towing’s case, FCP and CLS were already performing well, but LCP and Speed Index required optimization. ThatWare implemented several strategies to address these issues, such as reducing JavaScript execution, compressing and optimizing images, implementing lazy loading for media, and enhancing server response times. These improvements allowed the site to load faster, reducing bounce rates and keeping users engaged.

Enhancing Site Architecture

A well-organized site structure ensures that both search engines and users can navigate the website effectively. ThatWare applied a flat site architecture, minimizing the number of clicks needed to reach important pages. This structure made it easier for search engines to crawl and index pages while helping visitors find essential services, such as heavy equipment transport or towing and recovery. A simplified hierarchy also supports internal linking strategies, spreading page authority effectively across the site.

Mobile Usability and Responsiveness

With mobile devices accounting for a significant portion of search traffic, mobile usability is crucial. ThatWare ensured the Antonio’s Towing website was fully responsive, featuring readable text, appropriately sized images, and touch-friendly navigation. Mobile optimization improved user engagement, reduced bounce rates, and enhanced search engine rankings, as Google prioritizes mobile-first indexing for all websites.

Page Speed and Performance Optimization

Page speed directly impacts user experience and search engine ranking. Slow-loading pages can frustrate visitors and increase bounce rates, particularly for urgent service queries like towing. ThatWare optimized performance by enabling caching, streamlining HTML, CSS, and JavaScript files, and minimizing server response times. The result was faster page loads, a more enjoyable user experience, and improved Core Web Vitals scores.

Continuous Monitoring and Optimization

Technical SEO is an ongoing process. ThatWare established monitoring protocols using tools like Google Search Console, PageSpeed Insights, and performance analytics. Continuous monitoring allows the team to quickly identify new technical issues, track page performance, and make data-driven improvements over time. This proactive approach ensures the website remains optimized for both search engines and users as technology and algorithms evolve.

Impact on SEO and User Experience

By implementing these technical and UX enhancements, Antonio’s Towing experienced a noticeable improvement in both rankings and visitor engagement. Pages loaded faster, navigation became more intuitive, and users were able to access services quickly and efficiently. From an SEO perspective, search engines could crawl the site more effectively, resulting in better visibility for key service pages and higher organic traffic over time.

Content Refresh & Optimization

The Importance of Fresh, Relevant Content

Content is the backbone of SEO, and keeping it updated is essential for ranking and user engagement. For Antonio’s Towing, existing pages contained basic service information but lacked depth and actionable insights for users. ThatWare focused on refreshing content to better address the needs and search intent of potential customers. Updated content ensures that search engines recognize the site as authoritative and relevant, while users receive clear, helpful information about the services offered.

Updating Existing Pages

ThatWare began by reviewing all existing pages and identifying areas that required improvement. Service pages were enhanced with detailed descriptions, specific examples, and targeted keywords such as “heavy equipment transport Michigan” or “low boy equipment towing.” The content was structured using headings, bullet points, and concise paragraphs to improve readability and scanning. Internal linking was added to guide visitors to related services and important landing pages, increasing engagement and dwell time.

Creating New Content for Untapped Opportunities

New content was developed to address previously missed search queries and topics that were relevant to the target audience. Blog posts, how-to guides, and step-by-step tutorials were created to answer questions such as “how to prepare equipment for transport” or “safety tips for heavy towing.” This type of content not only attracts organic traffic but also positions Antonio’s Towing as a trusted resource in its industry.

Incorporating Visual Elements

Visual content such as infographics, diagrams, and images was added to enhance user engagement. Optimized images with descriptive alt text improved accessibility and provided additional opportunities for search engine indexing. Visual elements also make the content more engaging, helping to retain visitors and improve overall experience metrics.

Keyword Optimization and Intent Alignment

All refreshed and new content was optimized for search intent, ensuring alignment with user queries. ThatWare strategically incorporated primary and secondary keywords without overstuffing, focusing on natural language and readability. Meta descriptions, title tags, and headings were updated to reflect the content accurately and improve click-through rates from search engine results.

Measurable Impact of Content Refresh

The combination of updated service pages, new content, and optimized visuals led to significant improvements in engagement metrics. Users were able to find answers to their questions more easily, resulting in longer sessions, reduced bounce rates, and higher conversions. From an SEO perspective, refreshed content increased the site’s relevance for targeted keywords, contributing to higher rankings and organic visibility.

Off-Page SEO & Backlink Strategy

Understanding Off-Page SEO

Off-page SEO focuses on building authority and trust for a website through external factors, primarily backlinks. Backlinks from reputable websites signal to search engines that a site is credible, relevant, and valuable to users. For Antonio’s Towing, the initial backlink profile was limited, which constrained search engine rankings despite improvements in on-page SEO and technical performance.

Assessing the Backlink Profile

ThatWare conducted a thorough analysis of the existing backlink profile, examining the number of backlinks, domain authority of linking sites, relevance, and potential spam signals. At the start, Antonio’s Towing had a modest number of backlinks, with only a small percentage considered high-quality. This highlighted the need for a targeted strategy to acquire authoritative links that would improve the site’s overall domain authority and SERP competitiveness.

Building High-Quality Backlinks

The team implemented a strategic link-building campaign to acquire backlinks from relevant and authoritative sources. Tactics included:

• Guest posts on industry blogs and trade publications
• Partnerships with local business directories
• Listings on relevant niche websites related to towing, recovery, and heavy equipment transport
• Press releases highlighting business milestones or services

High-quality backlinks helped search engines recognize the site as trustworthy, improving rankings for key service pages and competitive keywords.

Focus on Relevance and Trust

It was not enough to build backlinks in quantity; quality and relevance were critical. ThatWare ensured all backlinks came from websites related to the towing, transport, or logistics industries. Avoiding spammy or low-authority links reduced the risk of penalties and strengthened the website’s reputation with search engines.

Monitoring and Refining Link Strategy

Backlink acquisition is an ongoing process. ThatWare set up monitoring systems to track new backlinks, assess their quality, and identify opportunities for additional link-building. Regular audits ensured the site maintained a healthy backlink profile, which is essential for sustaining long-term SEO success.

Impact on Domain Authority and Rankings

As a result of the backlink strategy, Antonio’s Towing saw measurable improvements in off-page SEO metrics. The number of referring domains increased, domain authority improved, and search engine rankings for target keywords became more competitive. Combined with on-page and technical optimizations, off-page SEO played a critical role in enhancing organic traffic and online visibility.

Performance Metrics & Results

Organic Traffic Growth

One of the most significant outcomes of the SEO campaign for Antonio’s Towing was the increase in organic traffic. After implementing technical optimizations, content refreshes, and search intent alignment, organic sessions increased by 125%, reaching 177 monthly visits. This growth represents a substantial reduction in dependence on paid search campaigns, signaling the effectiveness of a holistic SEO approach. Increased traffic also means more opportunities for lead generation, inquiries, and conversions, which directly impacts business revenue.

SERP Exposure and Rankings

The campaign led to a notable improvement in SERP visibility. The site began appearing for multiple high-intent keywords, achieving a top-10 average ranking for targeted searches like “heavy equipment transport Michigan” and “low boy towing services.” Increased impressions reflect the site’s broader visibility in search results, while the improved average position demonstrates the success of content alignment with user intent and keyword targeting.

Click-Through Rate (CTR) Improvement

The average CTR rose to 7.4% in the early stages of the campaign. This metric shows that not only were the pages ranking, but the content and meta descriptions were compelling enough to encourage users to click through to the website. Optimizing meta titles, descriptions, and headings for clarity and relevance played a key role in boosting engagement from search results.

Technical Performance Metrics

Technical improvements directly influenced both user experience and SEO performance. Core Web Vitals, including LCP, Speed Index, FCP, TBT, and CLS, were monitored and optimized. For instance:

• LCP: Reduced load time for the largest content element, enhancing perceived speed.
• Speed Index: Optimized to improve the visual loading experience.
• FCP & TBT: Streamlined JavaScript and server response times to reduce delays.
• CLS: Ensured layout stability to prevent disruptive content shifts.

Optimizing these metrics led to faster page loads, smoother navigation, and higher user satisfaction, which contributed to improved search engine rankings.

Crawl Stats and Indexing

Google Search Console data showed 238 crawl requests and an average response time of 587 milliseconds. These metrics indicate that search engines could crawl and index the site efficiently, ensuring that optimized content and new pages were discoverable. Reducing crawl depth and improving site structure contributed to this efficiency.

Domain and Page Authority Metrics

Off-page SEO efforts resulted in improvements in domain and page authority. The site acquired new backlinks, increasing the number of referring domains and boosting off-page scores. The spam score remained low at 1%, indicating high-quality link acquisition without risk of penalties. Moz Trust and authority metrics demonstrated steady growth, enhancing credibility in search engines’ eyes.

Search Performance Trends and Early Gains

Google Search Console performance data for Antonio’s Towing shows a substantial improvement over the past three months compared to the previous period. Total clicks increased from 0 to 38, and impressions rose from 0 to 517, reflecting growing visibility in search results. The average click-through rate (CTR) reached 7.4%, while the site achieved an average position of 9.7, signaling early success in targeting high-intent keywords. These metrics demonstrate that the SEO strategies implemented are beginning to drive meaningful traffic and engagement.

Visual Representation of Results

To convey progress effectively, the campaign results can be visualized using:

• Before-and-after charts showing organic traffic growth.
• Tables comparing Core Web Vitals pre- and post-optimization.
• Graphs tracking keyword rankings and CTR over time.
  These visualizations clearly highlight the tangible impact of a structured SEO strategy.

Analysis: What Worked and Why

The results show that a combination of search intent optimization, technical SEO improvements, content refresh, and quality backlink acquisition worked synergistically. Search engines rewarded pages that were technically sound, user-friendly, and aligned with user intent. The strategic combination of on-page and off-page efforts ensured sustainable growth rather than temporary spikes, establishing a foundation for ongoing SEO success.

Desktop Performance and Core Web Vitals Analysis

The PageSpeed Insights report for Antonio’s Towing’s desktop site shows an overall performance score of 70, indicating room for improvement. Key Core Web Vitals reveal strong metrics for First Contentful Paint (0.8 seconds), Total Blocking Time (60 ms), and Cumulative Layout Shift (0), all in the green and signaling good initial load stability and responsiveness. However, the Speed Index and Largest Contentful Paint both register 3.9 seconds, which is flagged as needing improvement, suggesting that while the site loads its first elements quickly, the largest visible content takes longer to appear, impacting perceived performance and overall user experience.

Domain Authority and Backlink Profile Analysis

The domain and page authority metrics for Antonio’s Towing reveal a developing but promising online presence. The website currently has a Domain Authority (DA) of 4 and a Page Authority (PA) of 16, supported by a total of 40 backlinks, of which 5 are high-quality, representing 13% of the profile. Moz Trust is measured at 2, the spam score remains low at 1%, and the off-page score stands at 30%, reflecting a solid foundation for future link-building efforts. At just 195 days old, the site is in its early stages of growth, and these metrics highlight both its current performance and the potential for establishing stronger authority over time.

Lessons Learned & Actionable Takeaways

Key Strategies That Drove Success

Several strategies were critical to the success of Antonio’s Towing’s SEO campaign:

• Aligning content with search intent to meet user needs.
• Optimizing technical SEO to improve Core Web Vitals and site crawlability.
• Refreshing and expanding content to cover missed queries.
• Building high-quality, relevant backlinks to increase authority.

Advice for Small Businesses in Competitive Niches

Small businesses often face challenges when competing with established players, but these lessons are universally applicable:

• Start with Search Intent Alignment: Understanding what your audience searches for ensures your content is relevant and valuable.
• Technical SEO is Not Optional: Fast, stable, and crawlable websites perform better in search engines and improve user experience.
• Content Refresh is Ongoing: Continuously updating and expanding content keeps your site relevant and authoritative.
• Authority and Link-Building are Critical: Quality backlinks improve domain authority and help you compete effectively for high-value keywords.

Recommendations for Continuous SEO Growth

Sustainable SEO growth requires ongoing attention and adaptation:

• Conduct regular audits of Core Web Vitals and technical performance.
• Monitor keyword rankings and adjust content strategies accordingly.
• Expand content targeting niche-specific queries and local searches.
• Continue building partnerships and authoritative backlinks for off-page strength.
• Track metrics like CTR, impressions, and conversions to identify areas for improvement.

GEO

How ThatWare Helped The Cambria International School Improve AI Visibility Through GEO 

The search landscape is evolving faster than ever before. Traditional SEO alone is no longer enough for brands that want to dominate visibility in modern AI-driven search ecosystems. Today, users are increasingly relying on AI platforms such as ChatGPT, Claude, Gemini, and Google AI Overviews to discover businesses, educational institutions, products, and services. Instead of browsing through multiple blue links, users now expect instant answers generated directly by artificial intelligence systems.

This transformation has created a new digital battleground where brands must optimize not only for search engines but also for AI-generated responses. As a result, Generative Engine Optimization Services are becoming a critical part of modern digital marketing strategies.

For educational institutions, this shift is even more important. Parents and students searching for the “best international schools,” “top junior colleges,” or “top schools near me” are increasingly receiving AI-generated recommendations before they even click on a website. If a school is not recognized, interpreted, and recommended by AI systems, it risks losing significant visibility and prospective admissions opportunities.

This is exactly where ThatWare stepped in to help The Cambria International School strengthen its digital footprint across AI-powered search environments.

Through advanced GEO strategies, semantic optimization, AI-focused content structuring, and intelligent search visibility enhancement, we helped the institution gain meaningful presence across multiple AI systems including ChatGPT, Claude, Gemini, and Google AI Overviews. The campaign focused not just on improving rankings, but on ensuring that AI engines recognized The Cambria International School as a trusted educational entity worthy of recommendation in AI-generated answers.

This case study explains how we implemented a strategic AI visibility campaign and how our GEO-driven methodologies transformed the school’s presence across modern AI search ecosystems.

Overview Of The Project

The Cambria International School is an international school and junior college focused on delivering world-class holistic education through internationally aligned curricula, experiential learning, modern infrastructure, and skill-based educational development.

The institution aims to prepare students for global opportunities and advanced academic pathways by integrating conceptual learning, modern teaching methodologies, and practical educational experiences. With a strong focus on student development, the institution has positioned itself as a forward-thinking educational organization within the Kalyan region.

However, despite offering quality educational services, the institution faced a major challenge in the emerging AI-search ecosystem.

Although the school maintained an online presence through its website, it lacked visibility for AI-generated search results and conversational search queries. The institution was not being effectively recommended by AI platforms when users searched for terms such as:

• Top international schools in Kalyan
• Best schools in Kalyan West
• Best junior colleges in Kalyan
• International schools near me
• IGCSE schools in Kalyan

As AI-powered discovery systems became more dominant, the absence of visibility across these platforms created a significant competitive disadvantage.

This is where our team initiated a complete GEO for Educational Institutions strategy designed specifically to improve AI-driven discoverability and entity recognition.

Advanced AI Infrastructure Implementation For GEO Growth at The Cambria International School

To strengthen AI visibility and improve discoverability across AI-powered search systems, ThatWare implemented a highly advanced GEO infrastructure layer on the website of The Cambria International School.

Modern AI systems such as ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews do not rely only on traditional SEO signals. These systems increasingly depend on semantic structures, machine-readable brand intelligence, entity mapping, contextual indexing, vector-based retrieval systems, and AI-guided crawling frameworks.

To address this, we implemented multiple AI-specific optimization files and semantic intelligence layers including:

• ai.txt
• llms.txt
• ai-manifesto.json
• vector-feed.xml
• semantic-sitemap.xml

These frameworks helped AI systems better understand the institution’s educational authority, curriculum relevance, geographical positioning, and semantic relationships.

1. AI.TXT Implementation

One of the first steps involved implementing an advanced ai.txt framework.

This file was specifically designed to provide direct AI crawling instructions for Large Language Models, retrieval engines, answer engines, and generative AI systems.

The file included:

• AI crawling permissions
• Entity-level guidance
• Educational contextual reinforcement
• Retrieval optimization instructions
• Conversational AI directives
• AI indexing priorities
• Brand attribution rules

Example AI.TXT Structure

Why This Was Important

Traditional robots.txt files are mainly designed for search engine crawlers. However, AI systems require additional semantic instructions that help them understand:

• What the brand represents
• Which topics are authoritative
• How educational entities should be interpreted
• Which contextual relationships must be preserved

This AI.txt framework improved machine-level comprehension across generative search systems.

2. LLMS.TXT Implementation

We also implemented a dedicated llms.txt framework.

This file acts as a semantic instruction layer specifically for Large Language Models.

It provides structured information about:

• The institution
• Educational services
• Curriculum categories
• Entity relationships
• AI usage permissions
• Knowledge graph associations

Example LLMS.TXT Structure

Why LLMS.TXT Matters

LLM systems retrieve information differently from search engines.

Instead of simply indexing pages, they build contextual relationships between:

• Educational services
• Location relevance
• Curriculum specialization
• Brand authority
• User intent

The llms.txt file helped reinforce these semantic relationships while improving contextual retrieval performance.

3. AI-MANIFESTO.JSON Implementation

One of the most advanced layers implemented during the campaign was the ai-manifesto.json framework.

This machine-readable semantic intelligence file provides AI systems with structured brand identity, entity definitions, contextual associations, topical authority, and semantic interpretation rules.

Example AI-MANIFESTO.JSON Structure

Strategic Impact Of AI-MANIFESTO.JSON

This framework helped AI systems:

• Understand the institution’s identity
• Associate the brand with international education
• Build educational entity relationships
• Strengthen semantic trust signals
• Improve retrieval consistency

It also enhanced AI interpretation accuracy during conversational search generation.

4. VECTOR-FEED.XML Implementation

Another critical implementation was the vector-feed.xml framework.

This file acts as a semantic vector indexing layer for AI retrieval systems.

Unlike traditional sitemaps, vector feeds help AI systems prioritize contextual understanding, semantic chunking, entity mapping, and vector embedding relevance.

Example VECTOR-FEED.XML Structure

Why VECTOR-FEED.XML Was Important

AI systems increasingly use vector embeddings and semantic chunking for information retrieval.

The vector-feed.xml framework helped:

• Improve semantic indexing
• Strengthen topic relationships
• Enhance contextual retrieval
• Improve AI recommendation accuracy
• Increase educational entity prominence

This significantly improved AI visibility within conversational search ecosystems.

5. SEMANTIC-SITEMAP.XML Implementation

We also implemented an advanced semantic-sitemap.xml framework.

Unlike traditional XML sitemaps, semantic sitemaps provide AI systems with contextual importance and semantic hierarchy.

Example SEMANTIC-SITEMAP.XML Structure

Strategic Advantages Of SEMANTIC-SITEMAP.XML

This implementation helped AI systems:

• Understand content hierarchy
• Prioritize educational pages
• Improve semantic categorization
• Associate pages with educational intent
• Improve contextual indexing

As a result, the institution gained stronger visibility across AI-generated educational recommendations.

How These AI Frameworks Improved GEO Growth

The implementation of these AI-focused frameworks created a highly optimized semantic ecosystem around The Cambria International School.

The combined impact included:

• Improved AI crawling efficiency
• Better semantic understanding
• Stronger entity recognition
• Enhanced contextual retrieval
• Increased conversational visibility
• Higher recommendation probability across AI systems

After implementing these GEO infrastructure layers, the institution achieved visibility improvements across:

• ChatGPT recommendations
• Claude educational results
• Gemini AI search responses
• Google AI Overviews

The school began appearing prominently for searches such as:

• Top International School in Kalyan
• Top Schools in Kalyan West
• Best Junior Colleges in Kalyan

This demonstrated how advanced GEO infrastructure can significantly strengthen AI visibility and semantic authority within modern AI-powered search ecosystems.

Results After GEO Implementation

The implementation of our GEO strategy generated substantial improvements in AI visibility across multiple AI platforms.

Most importantly, the institution began appearing directly within AI-generated recommendations for highly relevant educational searches.

These results demonstrated that AI systems had started recognizing The Cambria International School as a trusted educational entity.

ChatGPT Presence Improvements

One of the biggest successes of the campaign was the institution’s improved visibility within ChatGPT-generated recommendations.

For the search query:

“Top International School in Kalyan”

The institution appeared in the 3rd position within ChatGPT recommendations.

This was a major milestone because ChatGPT-generated recommendations rely heavily on semantic trust, entity relevance, contextual authority, and informational consistency.

The recommendation highlighted:

• Modern globally focused education
• Conceptual learning
• Experiential learning methodologies

This demonstrated that the AI system had developed a contextual understanding of the institution’s educational positioning.

For the search query:

“Top Schools in Kalyan West”

The institution appeared in the 4th position within ChatGPT recommendations.

This visibility indicated strong progress in conversational educational search optimization.

The AI system recognized the institution for:

• Modern facilities
• Smart classrooms
• Quality teaching methodologies
• Local educational reputation

These results validated the effectiveness of our Generative Engine Optimization Services strategy in improving conversational AI visibility.

Claude Presence Improvements

The campaign also generated strong visibility improvements within Claude AI search results.

For the search query:

“Top International School in Kalyan”

The institution appeared in the 1st position within Claude-generated recommendations.

This was one of the strongest indicators of successful semantic authority development.

Claude identified the institution as:

• A quality IGCSE educational institution
• A school known for modern learning approaches
• A globally aligned educational organization

Achieving the top position demonstrated that the institution had become highly recognizable within AI-generated educational searches.

For the query:

“Top Schools in Kalyan West”

The institution appeared within the Top 10 recommendations generated by Claude.

This further strengthened the institution’s authority within localized educational searches.

Gemini Presence Improvements

Google Gemini also showed major improvements in educational recommendation visibility.

For the query:

“Top International School in Kalyan”

The institution appeared in the 1st position within Gemini recommendations.

Gemini recognized the institution as:

• A premier choice for international education
• A provider of globally aligned curriculum
• A leading educational institution in Kalyan West

This result demonstrated strong semantic authority and successful AI retrieval optimization.

For the query:

“Top Schools in Kalyan West”

The institution appeared in the 6th position within Gemini-generated recommendations.

This visibility further reinforced the institution’s educational authority across localized AI search ecosystems.

Google AI Overview Presence

One of the most impactful achievements of the campaign was the institution’s appearance within Google AI Overviews.

The institution appeared within AI Overview results for:

“Top International School in Kalyan”

Google AI Overview specifically mentioned the institution among prominent international schools in the region.

This represented a major visibility milestone because AI Overviews increasingly dominate informational search experiences within Google.

The institution also appeared within AI Overview results for:

“Best Junior Colleges in Kalyan”

Google highlighted the institution as one of the highly rated junior colleges in the region.

This significantly expanded visibility beyond international school-related searches and strengthened educational category relevance.

The success of these placements clearly demonstrated the impact of strategic Generative Engine Optimization Services combined with semantic authority building and AI visibility enhancement.

LLM

How ThatWare’s LLM Optimization Strategy Increased PurpleGlo’s Visibility across AI Search Ecosystems?

Today, users increasingly rely on AI-powered platforms to discover agencies, compare service providers, and find trusted recommendations. Instead of browsing multiple websites, users ask conversational questions and expect direct answers from AI systems. This shift has created a new optimization frontier powered by LLM Optimization.

As a leading digital experiential agency, PurpleGlo delivers innovative experiences that combine technology, creativity, and audience engagement. However, despite its expertise and strong portfolio, the website lacked discoverability across AI-powered search environments.

The challenge was not service quality. The challenge was visibility. To address this challenge, ThatWare developed a comprehensive AI Search Optimization strategy focused on improving visibility across modern answer engines and AI-driven search ecosystems.

The results became highly impactful.

Following implementation, PurpleGlo achieved:

• First-position visibility on ChatGPT for “top digital experiential agency in UAE”
• First-position visibility on ChatGPT for “best interactive book projection agency in UAE”
• First-position visibility on Gemini for both target queries
• Fourth-position visibility on Claude for both target queries
• Presence within Google AI Overview results for relevant service-based searches

These achievements demonstrated how a carefully structured AI Visibility Strategy can strengthen discoverability across emerging search ecosystems.

This case study explores how ThatWare helped PurpleGlo improve visibility through Answer Engine Optimization, Generative Engine Optimization, semantic search optimization, and AI-first content strategies.

Implementing GEO to Improve Discoverability across AI Search Platforms

A major component of this campaign involved Generative Engine Optimization. Modern AI systems generate answers differently from traditional search engines.

Instead of matching keywords alone, they evaluate contextual relevance, expertise signals, and semantic relationships before recommending businesses.

ThatWare implemented GEO-focused improvements designed to increase PurpleGlo’s discoverability within AI-generated responses.

The campaign focused on:

• Semantic content enhancement
• Contextual authority building
• Entity optimization
• Search intent alignment
• Conversational query relevance

Optimizing for High-Intent Conversational Queries

The campaign also focused on the specific queries users were asking AI platforms. Research identified strong opportunities around searches such as:

• Top digital experiential agency in UAE
• Best interactive book projection agency in UAE
• Event experience solutions provider
• Experiential marketing agency UAE

Instead of targeting these queries through conventional keyword placement alone, ThatWare optimized content around the intent behind each search. This approach improved compatibility with AI systems that prioritize contextual understanding.

ChatGPT Visibility Results

One of the strongest outcomes of the campaign came from ChatGPT visibility.

Following the implementation of the LLM optimization strategy, PurpleGlo achieved first-position visibility on ChatGPT for: Top digital experiential agency in UAE

The website also secured first-position visibility for: Best interactive book projection agency in UAE

These results demonstrated that the implemented Answer Engine Optimization framework successfully improved PurpleGlo’s discoverability within conversational AI environments.

Claude Visibility Results

The campaign also generated visibility improvements within Claude. Although the rankings differed from ChatGPT, the visibility confirmed that the optimization framework was successfully influencing multiple AI ecosystems.

This reinforced the effectiveness of ThatWare’s broader AI Search Optimization strategy.

Gemini Visibility Results

PurpleGlo achieved particularly strong results within Gemini. The campaign secured first-position visibility for:

These outcomes demonstrated strong contextual relevance and improved entity recognition within Google’s AI ecosystem.

The results also validated the effectiveness of combining LLM Optimization, AEO, and GEO under a unified visibility strategy.

Google AI Overview Visibility

The campaign also generated visibility within Google AI Overview results.

This achievement is particularly important because AI Overview increasingly influences how users discover information within Google’s search experience.

This visibility expanded the brand’s presence beyond standalone AI assistants and into Google’s AI-generated answer ecosystem.

Measuring The Campaign’s AI Visibility Impact

The success of this campaign can be measured through its visibility improvements across major AI platforms. Following the implementation of ThatWare’s LLM Optimization framework, PurpleGlo achieved strong placement for highly relevant commercial-intent queries.

These results demonstrated the effectiveness of combining LLM Optimization, AEO, and GEO within a single visibility framework.

The campaign successfully expanded PurpleGlo’s discoverability across multiple AI-driven search environments.

AEO

How ThatWare’s Answer Engine Optimization Helped Safety House General Trading LLC Improve AI Visibility

The way users search for products and suppliers has changed dramatically. Traditional SEO is no longer enough to remain visible in modern search ecosystems.

Today, buyers increasingly rely on conversational AI platforms and answer engines to discover suppliers, compare providers, and make purchase decisions. This shift has made Answer Engine Optimization (AEO) a major factor in sustainable digital visibility.

Safety House General Trading LLC faced a similar challenge. Despite offering road safety and industrial safety products, the website lacked visibility across AI-driven search environments.

Users were actively searching for suppliers using high-intent queries like “wheel chock supplier in UAE”, but the website was not appearing prominently in conversational search platforms.

The challenge was not product demand. The challenge was discoverability in evolving AI-powered search systems.

ThatWare introduced a structured AEO service campaign focused on search intent optimization, semantic content structuring, conversational discoverability, and AI search optimization.

The results became highly noticeable across major AI platforms.

Safety House General Trading LLC achieved:

• 1st position visibility on ChatGPT for “wheel chock supplier in UAE”
• 1st position visibility on Perplexity for “top wheel chock supplier in UAE”
• 1st position visibility on Gemini for “top wheel chock supplier in UAE”
• 4th position visibility on Claude for “wheel chock supplier in UAE”

These outcomes demonstrated how a structured AI visibility strategy can strengthen discoverability beyond traditional search engines.

This SEO case study explores how ThatWare helped Safety House General Trading LLC improve discoverability through AEO, semantic search alignment, and conversational AI visibility strategies.

Let’s go into the details.

Why Answer Engine Optimization Became Critical for This Campaign?

The digital search ecosystem is rapidly evolving toward conversational discovery.

Today, industrial buyers no longer rely only on traditional search engines to find suppliers. Many now ask AI assistants direct questions to receive recommendations instantly.

Instead of typing broad keywords in ChatGPT, users ask queries like:

“Who is the best wheel chock supplier in UAE?”

This shift made Answer Engine Optimization a critical part of Safety House General Trading LLC’s growth strategy.

ThatWare optimized the website for conversational AI discovery using a structured framework focused on semantic understanding, contextual authority, and answer-focused discoverability.

Several key strategies were implemented:

• Conversational query optimization
• FAQ-focused content structure
• Supplier-intent content optimization
• AI-readable semantic architecture
• Contextual content mapping
• GEO-targeted discoverability improvements

These implementations helped conversational AI systems understand the website’s contextual relevance more effectively. As a result, Safety House General Trading LLC gained strong visibility across major answer engines and AI-powered search systems like ChatGPT.

Claude Visibility Building Discoverability across Multiple AI Platforms

The campaign also strengthened visibility on Claude, another emerging conversational AI platform used for research and recommendations.

Safety House General Trading LLC achieved the 4th position on Claude for the query:

“wheel chock supplier in UAE”

Although the positioning differed from other platforms, this visibility still demonstrated successful AI discoverability across multiple ecosystems.

This multi-platform visibility approach is important because different users rely on different AI systems during supplier research. Instead of optimizing for one platform alone, ThatWare focused on broader AI-powered search visibility.

Getting Visibility Regarding High-Intent Supplier Searches on Perplexity

Perplexity has become one of the fastest-growing answer engines for research-focused users.

The campaign achieved a major milestone when Safety House General Trading LLC appeared in the 1st position on Perplexity for:

“top wheel chock supplier in UAE”

This outcome highlighted stronger semantic authority and contextual relevance for supplier-related searches.

Perplexity prioritizes trusted and highly contextual information. Therefore, appearing prominently reinforced the website’s supplier credibility and discoverability.

The result also showed how Generative Engine Optimization strengthens visibility in emerging AI ecosystems.

Gemini Visibility Expanding Discoverability Across Google’s AI Ecosystem

The campaign also achieved strong discoverability within Gemini, Google’s AI-powered assistant platform.

Safety House General Trading LLC achieved the 1st position on Gemini for:

“top wheel chock supplier in UAE”

This achievement became especially valuable because Gemini integrates closely with Google’s search intelligence ecosystem. Strong visibility here improved trust signals and expanded brand discoverability through conversational recommendation systems.

The result reinforced the effectiveness of ThatWare’s AI search optimization strategy for supplier-based businesses.

AI Visibility Infrastructure Implemented by ThatWare for Safety House General Trading LLC

To improve semantic discoverability and strengthen visibility across conversational AI systems, ThatWare implemented an advanced AI visibility infrastructure for Safety House General Trading LLC.

These implementations helped answer engines, large language models, and AI retrieval systems understand the business, products, and supplier relevance more effectively.

As a result, the website gained stronger contextual understanding across modern AI-powered search ecosystems.

ai.txt

ThatWare implemented an AI crawler governance layer through ai.txt. This framework helps guide AI crawlers and answer engines toward priority website content while reducing crawl inefficiencies.

It supports discoverability for systems, including:

• GPTBot
• ChatGPT-User
• Google-Extended
• PerplexityBot
• Anthropic AI

This implementation strengthened AI-powered search visibility across conversational systems.

llms.txt

ThatWare created an AI accessibility framework through llms.txt. This structure helps large language models understand:

• Website hierarchy
• Product taxonomy
• Supplier positioning
• Brand relevance
• Content accessibility

This improved semantic understanding across answer engines and conversational retrieval systems.

ai-manifesto.json

ThatWare implemented ai-manifesto.json to improve machine-readable brand understanding.

This structure helps AI systems understand:

• Business identity
• Supplier positioning
• Product intelligence
• Entity relationships
• Trust and contextual signals

As a result, conversational systems can interpret Safety House General Trading LLC more accurately within industrial supplier-related queries.

vector-feed.xml

ThatWare implemented vector-feed.xml to improve semantic retrieval and contextual indexing.

This structure helps:

• Vector databases
• Retrieval systems
• Generative AI environments
• Semantic search platforms

understand relationships between products, supplier context, and business relevance more effectively.

This became an important layer of the campaign’s conversational search optimization strategy.

semantic-sitemap.xml

ThatWare also implemented semantic-sitemap.xml to strengthen machine readability and contextual indexing.

Unlike traditional XML sitemaps, this framework prioritizes:

• Semantic relationships
• Contextual understanding
• Entity mapping
• AI-first crawling signals

This improved discoverability across answer engines and semantic retrieval ecosystems.

How The Campaign Redefined Supplier Discoverability through AI?

The Safety House General Trading LLC campaign highlights how modern supplier visibility now extends far beyond traditional search rankings.

By combining Answer Engine Optimization, semantic indexing, conversational discoverability, and AI infrastructure implementation, ThatWare helped the business strengthen visibility across multiple answer engines.

The campaign successfully transformed Safety House General Trading LLC into a more discoverable supplier across:

• ChatGPT
• Claude
• Perplexity
• Gemini

This approach created stronger contextual authority and sustainable organic visibility growth across evolving AI-driven search ecosystems.

Commercial-intent keywords: optimize for decisions, not just definitions

A major opportunity for ThatWare’s own article and for client content programs is to connect educational questions to the services buyers are actively evaluating. Search volume alone is a poor filter for high-value service topics. A question asked by ten qualified procurement teams can be more valuable than a definition queried by thousands of casual readers.

We organize commercial intent into four clusters. Problem-aware queries include “why is my brand missing from ChatGPT recommendations” and “AI Overview cites competitors instead of us.” Solution-aware queries include “AI visibility audit services,” “AI search optimization services” and “generative engine optimization services.” Provider comparison queries include “enterprise GEO agency,” “GEO consultant vs SEO agency” and “best AI search visibility agency for B2B SaaS.” Measurement queries include “AI recommendation tracking,” “AEO reporting for marketing teams” and “AI citation audit.”

The content strategy should serve each stage honestly. A problem-aware page should diagnose causes before selling a service. A solution-aware page should explain scope and deliverables. A comparison page should outline evaluation criteria. A measurement page should define its methodology and limitations. This gives a brand an actual information architecture rather than a collection of keyword variants pointing to the same thin proposition.

Intent cluster	Representative commercial keyword	Best page or asset	Buyer question answered
Diagnosis	AI visibility audit services	Diagnostic service page + sample audit	What is missing and how will we identify it?
Provider selection	Generative engine optimization agency	Service methodology + proof	Who can manage a cross-platform GEO program?
Implementation	AEO and GEO services	Work plan + deliverable examples	What changes will be made to our site and evidence?
Reporting	AI recommendation tracking	Measurement methodology	How do we know whether the brand is selected?
Entity clarity	Vector Entity Modelling services	Framework + use cases	Why is our company described incorrectly?
Enterprise rollout	Enterprise AI search optimization	Governance and rollout model	Can this operate across products, regions and teams?

For this article specifically, the most natural internal path is from the educational framework to relevant service and methodology pages. Readers can explore ThatWare’s AVM and VEM framework, the Vector Entity Modelling methodology and the five-layer GEO stack when they want more depth. Internal links should advance the reader’s job rather than function as repetitive keyword anchors.

A five-part AI visibility audit that leads to action

The audit is where the idea becomes a working business process. A high-quality AI visibility audit should deliver a reproducible baseline, a diagnosis and a prioritized fix plan. It should not be a folder of impressive screenshots without enough detail to test the conclusions later.

Part 1: define the question universe

Build prompts around real search and buying behavior. Include informational questions, comparison requests, local needs, commercial qualification, alternative-provider searches and brand-specific verification. Separate branded prompts from nonbranded prompts; otherwise a dashboard can look healthy simply because it mostly asks an assistant about a company by name.

Agree on the industries, languages, markets and competitors in scope. Write down why each prompt matters. A security buyer’s question about data processing is different from a founder’s question about agency pricing. One prompt library cannot represent every buyer without intentional segmentation.

Part 2: capture actual answers and supporting sources

Run an agreed sample across relevant AI-assisted search and answer products. For each capture, preserve the prompt, response, source URLs where available, timestamp, platform and important settings. If repeatability is being assessed, use multiple runs and paraphrases. Where a product does not offer citations, do not manufacture them in the report.

Collection methods must respect each provider’s terms and technical limits. Manual sampling, approved APIs or suitable compliant tools can be used, but the methodology should be documented. A claim of “AI market share” without a clear sample definition is not an actionable business metric.

Part 3: map citations to exact claims

Identify which domains and pages support the generated answer. Then inspect the specific statements made about the brand. Does a citation support the asserted price, feature, credential or service location? Are important qualifications missing? Is a competitor cited for a claim that your own page establishes more clearly? Is the assistant repeating outdated brand information?

This separates citation frequency from citation fidelity. It also generates specific content tasks, such as updating a product matrix, making evidence visible in HTML or clarifying an ambiguous service description.

Part 4: score the commercial conclusion

Record whether the brand was absent, mentioned, explained, compared or recommended. Where a shortlist exists, note the position and the stated rationale. Record inaccuracies, unsupported praise, inappropriate recommendations and relevant competitors. Positive language alone is not sufficient if the system has misunderstood the buyer’s actual requirement.

A conclusion score should reflect the prompt and the use case. A local service provider recommended outside its service area should not be counted as a quality win. A brand correctly omitted from a category it does not serve should not automatically be classified as failure.

Part 5: produce a prioritized implementation backlog

Turn gaps into owners and fixes. Typical outputs include a technical crawl list, canonical entity inventory, citation corrections, content briefs, comparison-page outlines, evidence collection requests, structured-data validation tasks and a retesting calendar. Rank tasks by commercial relevance, factual risk, implementation effort and confidence in the diagnostic evidence.

ThatWare’s audit model combines sampled answer analysis with an entity repair plan and a practical 90-day workstream. The objective is not a black-box score. It is a sequence of verifiable changes that both search teams and commercial stakeholders understand.

Figure 12. ThatWare’s audit framework links prompt sampling, answer capture, source mapping, conclusion assessment and an actionable repair backlog. Credit: ThatWare brightonSEO presentation.

How to measure AI visibility without confusing activity with value

There is no single perfect KPI for AI-driven discovery. The measurement plan should connect what can be observed to the questions leadership actually asks. A marketing director wants to know whether the company is gaining qualified consideration. A content manager wants to know whether important evidence appears in answers. A sales leader wants to know whether new prospects arrive with accurate expectations.

We recommend separating leading indicators from business outcomes. Leading indicators include crawlable coverage, canonical entity consistency, source accuracy, answer presence and comparison inclusion. Outcomes include qualified inbound enquiries, pipeline influence where reasonably attributable, branded discovery trends and measurable conversion performance. The first group can change faster. The second can be influenced by many channels and should not be attributed casually to a single AI answer.

Metric	Working definition	Common interpretation mistake
Prompt-set presence	Share of sampled relevant prompts that mention the brand	Treating a small prompt set as the whole market
Citation rate	Share of sampled answers that cite an owned source	Treating citation as an endorsement
Claim fidelity	Share of checked material claims supported by their cited sources	Assuming a link proves the adjacent sentence
Comparison inclusion	Share of comparison prompts including the brand	Ignoring whether the comparison is accurate
Qualified recommendation	Share of relevant buying prompts recommending the brand appropriately	Rewarding an unsuitable or fabricated recommendation
Stability	Consistency across repeated prompts, models and time	Treating one screenshot as a reliable trend
Business impact	Qualified leads, pipeline signals or conversions with a documented path	Claiming causation from correlation alone

In addition to independent monitoring, Google now offers a directly relevant first-party signal. Its Search Console generative AI performance reports, rolled out globally by August 2026, provide visibility into website impressions in generative AI features, including pages, locations, devices and dates where available. These reports do not reveal a third-party tool’s imagined “Google AI trust score” or the full reasoning behind a recommendation. They are still useful as a verified part of the broader measurement picture.

Measurement should be disciplined about geography and wording. An assistant’s answer can shift when asked from another country, given a different budget, or asked whether a brand is suitable for a regulated market. We prefer a trend with its sample size, uncertainty and prompt definitions attached to a single headline percentage without context.

How ThatWare uses AVM and VEM to separate visibility from understanding

ThatWare’s proprietary AI Visibility Metric (AVM) and Vector Entity Modelling (VEM) frameworks address complementary questions. AVM concerns observed brand visibility across a defined sample of AI answers: presence, citations, position, consistency and surrounding evidence. VEM concerns the entity foundation behind those outcomes: brand clarity, content coverage, authority relationships, machine-readable organization and relevant query coverage.

The simple distinction is this: AVM asks whether AI-facing systems surface the brand; VEM asks how clearly the brand is represented and connected to its actual category and capabilities. A company can perform well on one dimension and poorly on the other. An established brand may be mentioned frequently but described using outdated information. A specialized provider may have a coherent and well-supported entity profile yet remain absent from broad comparison answers.

These are ThatWare-defined diagnostic frameworks, not metrics issued or endorsed by OpenAI, Google, Anthropic, Perplexity or another provider. The methods can help identify gaps and organize work, but an AVM or VEM result should always be interpreted alongside the prompts, sources, dates and scoring rules used in the assessment. We would rather have a transparent, imperfect diagnostic than an opaque number that implies control over proprietary AI systems.

In practice, AVM identifies where visibility is missing; VEM investigates whether the underlying entity representation is part of the problem. The combined output informs content planning, entity cleanup, evidence acquisition, navigation, structured information and ongoing monitoring. Readers seeking the methodology can review ThatWare’s AVM and VEM explanation and VEM service framework.

Figure 13. AVM measures sampled AI visibility outcomes; VEM examines the brand’s entity structure and understanding. These are ThatWare frameworks, not official platform scores. Credit: ThatWare brightonSEO presentation.

What this looks like in three commercial markets

The same principle can produce very different priorities depending on the market. These are illustrative scenarios, not claimed results from named ThatWare clients.

Scenario A: enterprise software with strong rankings but weak shortlists

A software provider ranks well for “enterprise workflow automation” and publishes frequent industry articles. However, an assistant repeatedly cites its explainer material while recommending larger competitors. A review finds that the provider’s own pages lack a direct integrations matrix, clear deployment options, documented security controls and a practical comparison of ideal customer profiles.

The first fix should not be fifty more general-purpose posts. It should be decision-support information buyers need: supported integrations, technical constraints, deployment models, detailed product pages and appropriately approved proof. The audit can then retest integration-specific and enterprise procurement prompts to see whether the answers represent the provider more accurately.

Scenario B: a local professional service with inconsistent identity

A multi-location specialist is mentioned in broad explanations of its field but rarely appears in local recommendations. Office addresses, opening hours and service boundaries are inconsistent across its website and public profiles. Several service pages also use overlapping names, making it difficult to determine which location handles which request.

The plan centers on accurate local pages, verified profiles, unambiguous service areas, better internal links and updated qualifications. Where relevant, local business and organization markup should reflect visible content. The measurement prompts must be evaluated for legitimate geographic fit, not just whether the provider’s name appears.

Scenario C: a specialist manufacturer with strong technical evidence

A manufacturer has excellent downloadable specifications and product testing documents, but its HTML product pages are minimal. AI explanations occasionally use the PDF data; buying-stage comparisons omit the brand. Commercial buyers cannot quickly find exact use cases, available configurations, region-specific compliance information or distributor paths.

Here the opportunity is to expose essential verified facts on accessible product pages, connect the technical documentation to each relevant product, explain application boundaries and improve buyer-oriented comparison content. Technical depth becomes commercially visible when it is organized for both experts and decision-makers.

Across these scenarios, better AI visibility starts with better public evidence. The changes also improve sales enablement and customer experience, which is why this work should not sit in an isolated experimental SEO budget.

Inside an AI Visibility Audit: What ThatWare’s AVM and VEM Scores Reveal

A high organic ranking can tell us that a page is discoverable. It cannot tell us whether an AI assistant recognizes the company behind that page, uses its evidence, or considers it a strong recommendation when a buyer compares providers. Those are different outcomes, and they call for a different kind of audit.

At ThatWare, we developed two complementary diagnostic frameworks to examine that gap. AI Visibility Metric (AVM) evaluates a brand’s presence and treatment in a defined sample of AI-assisted discovery results. Vector Entity Modelling (VEM) looks at the digital identity, topical connections and supporting evidence that make the brand easier to interpret. Advanced AVM and Advanced VEM extend those assessments into deeper strategic indicators.

Figure 14. ThatWare’s AVM dashboard snapshot for an OpenAI-labeled assessment, showing an Emerging overall score and the sampled topic, industry and country scope. Image credit: ThatWare, supplied AVM report screenshot.

The following dashboard example makes the distinction concrete. It is an assessment of ThatWare itself, displayed in an OpenAI-labeled report and organized around professional SEO services and advanced SEO. The specified geographic scope includes India, the USA, the UAE, the UK and Australia, with English as the language. The service-intent phrases include professional seo services agency, advanced seo, advanced seo agency, professional seo companies and expert seo company.

Important: these scores come from ThatWare’s own scoped diagnostic, not an official OpenAI rating or a measurement of all ChatGPT answers. Interpreting them responsibly requires the underlying prompts, responses, citations and scoring method.

Reading the main AVM score: 35.86/100, with an Emerging classification

The first panel reports a 35.86/100 AVM score, categorized as Emerging. Its adjacent visibility display reads 36%. The display should not be interpreted as proof that 36% of AI search users saw ThatWare or that the brand appeared in exactly 36% of all possible responses.

The practical reading is more useful than the number alone: the brand has some recognizable footprint in the tested environment, but the selected evidence does not yet support consistently strong performance across the report’s dimensions. There is a difference between being detectable in some category-related answers and being the provider a system cites or selects when the question becomes commercial.

To understand where that difference arises, we need the breakdown rather than the headline score. The main AVM methodology considers five performance dimensions: presence, citation, authority, consistency and position. The dashboard also displays a confidence indicator, which should be read separately as information about the sufficiency or quality of the assessment evidence, not as another term for brand trust.

Figure 15. AVM breakdown for the same assessment: Presence 50.00; Citation 13.33; Authority 46.67; Consistency 41.17; Position 26.67; Confidence 60.90. Image credit: ThatWare, supplied dashboard screenshot.

What the six AVM breakdown indicators mean

Dashboard indicator	Reported result	Business interpretation
Presence	50.00/100 — Developing	The brand is recognizable in some sampled contexts, but presence alone says little about shortlist strength or purchase suitability.
Citation	13.33/100 — Weak	The evaluated answers provide limited qualifying citation support. This is the most conspicuous gap between being encountered and being used as a source.
Authority	46.67/100 — Developing	The report sees a partial credibility footprint, with room for stronger, verifiable owned and independent evidence.
Consistency	41.17/100 — Critical	The brand’s identity and topical associations need closer checking across sampled contexts. The score is not a literal percentage of incorrect facts.
Position	26.67/100 — Weak	Even where the brand is recognized, its placement within sampled answer or comparison structures is weaker than desired. This is not its Google ranking position.
Confidence	60.90/100 — Developing	The report has a separate evidence-confidence assessment. This is not a model provider’s certification of ThatWare’s credibility.

The contrast between Presence at 50.00 and Citation at 13.33 is the heart of this example. A brand can be recognized without its website, research or service pages becoming the evidence that an answer points to. For an SEO agency, this may look like an assistant understanding the company’s name but leaning on independent comparisons, directories or competing publishers to explain the category.

The low Position score of 26.67 raises a second question. When a prospect asks an assistant to compare advanced SEO agencies, is ThatWare presented as a relevant option, an incidental mention, or not included at all? A general category reference is not equivalent to a purchase-stage recommendation. The audit therefore needs to preserve the exact prompt, the answer context and any list ordering before interpreting the result.

Authority at 46.67 and Consistency at 41.17 offer two different lines of investigation. An authority review asks whether case studies, named expertise, published methods, external coverage and independent validation substantiate the brand’s claims. A consistency review asks whether the same company, services, leadership and positioning can be identified across owned pages and credible third-party sources. One does not automatically repair the other.

Finally, Confidence at 60.90 must not be used to cancel out the weak citation result. A dashboard can be reasonably confident that its observations reveal an issue and still report poor visibility within that sample. Decision-makers should ask how many prompts and runs supported the score, how they were selected, and how much uncertainty remains.

Advanced AVM: nine signals that reveal the quality of AI visibility

The Advanced AVM view adds a more strategic reading of the same broad problem. It looks beyond basic appearances to themes such as discovery strength, trust, repeatability and relative presence within the tested answer environment. These are modelled diagnostic indicators. They should not be presented as direct measurements of an AI provider’s hidden algorithm or a scientifically calibrated probability of recommendation.

Figure 16. Advanced AVM Intelligence panel showing nine strategic scores, including Discoverability, AI Trust, Answer Probability and AI Share of Voice. Image credit: ThatWare, supplied screenshot.

Advanced AVM indicator	Score /100	How to interpret it responsibly
AI Discoverability	43.00	A signal of how discoverable the brand appears in the assessed contexts, not a count of every retrieval event.
AI Trust	31.00	The assessment identifies a comparatively weak trust-support profile; the score is not an official trust grade issued by OpenAI.
Entity Dominance	39.00	The brand does not strongly own the sampled topic associations or category conversations.
Answer Probability	34.00	A model-derived readiness indicator, not a measured 34% chance of appearing in a consumer answer.
AI Volatility Stability	54.00	The strongest of these nine scores; it provides a relative stability signal within the assessed methodology, not guaranteed consistency.
AI Memory	40.00	A proxy for repeat recognition or durable association, not proof of permanent memory stored by the AI provider.
Entity Sentiment	47.00	A view of how favorable the brand’s assessed framing is, which must be checked against the exact answer text.
AI Market Share Visibility	28.00	A weak competitive visibility signal in the sample, not a measurement of actual commercial market share.
AI Share of Voice	33.00	A relative visibility measure for the tested query universe, not 33% of all AI conversations.

Three patterns stand out. First, discoverability is not converting cleanly into authority. AI Discoverability scores 43.00, while AI Trust is 31.00. An agency may have accessible pages and clear services while still lacking the independent, claim-specific evidence that makes it compelling in comparison answers. Publishing more generic service copy would not necessarily solve that problem.

Second, recognition is not the same as competitive ownership. Entity Dominance scores 39.00, while AI Market Share Visibility is 28.00 and AI Share of Voice is 33.00. These indicators prompt an important commercial investigation: which brands repeatedly occupy the strongest positions for terms such as advanced SEO agency or professional SEO company, what evidence supports them, and which specific types of queries produce the largest gap?

Third, a relatively stronger stability indicator does not establish a strong recommendation position. AI Volatility Stability is 54.00, higher than the other Advanced AVM measures. That may indicate the tested evidence is not changing wildly across the assessed conditions, but it is not a reason to overlook weak trust, citation and commercial visibility. Stable underperformance is still a problem worth addressing.

An agency evaluating AI search visibility optimization should use these advanced indicators as prompts for deeper analysis. It should not claim that a numerical score reveals the internal state of a proprietary model. The goal is to connect the score back to an observed response and then to a legitimate, verifiable content or authority improvement.

VEM goes below the answer: does the AI ecosystem understand the brand?

AVM tells us what the sampled outputs look like. VEM asks whether the brand’s digital evidence explains why those outputs may be strong, weak or inconsistent. It evaluates identity clarity, content coverage, authority, entity relationships, AI readiness and query coverage in a structured framework.

The screenshot reports an OpenAI Vector Entity Modelling score of 36.85/100, labelled Poor Entity Foundation. Its executive summary describes a recognizable brand with some established identity, but insufficient depth in third-party authority, citations and non-branded discovery signals for a stronger position against category competitors.

This is not a declaration that the brand is unknown, nor a permanent verdict on its reputation. It is a diagnostic signal within the selected professional SEO and advanced SEO topic set. The more useful insight is that recognition of a company name may coexist with weak connections to the specific buyer problems for which it wants to be recommended.

Figure 17. Vector Entity Modelling dashboard with a 36.85/100 score, explanatory executive summary and six core intelligence dimensions. Image credit: ThatWare, supplied screenshot.

VEM intelligence layer	Score /100	What the brand team should examine
Brand Intelligence	37.00	Canonical name, approved variations, service positioning, leadership and recognition of the correct business.
Content Intelligence	34.00	Depth and consistency of the pages that substantiate priority services, methods, use cases and comparison questions.
Authority Intelligence	41.00	Independent validation, credible mentions, documented expertise, citations and named evidence.
Entity Intelligence	39.00	Connections among the brand, people, services, methodologies, publications and relevant topics.
AI Readiness	36.00	Crawlable, accessible, well-structured content and reliable factual signals available to relevant systems.
Query Intelligence	32.00	Coverage of the non-branded, comparative and commercial questions that prospects actually ask.

Query Intelligence at 32.00 is especially relevant for a growth team. A company may be recognized when someone types its name, but the revenue opportunity often begins when the buyer has not decided whom to hire. Questions such as “Which agency can audit our brand in ChatGPT and Google AI Overviews?” or “Who offers advanced SEO and GEO consulting for an international B2B business?” test commercial relevance rather than brand recall.

Content Intelligence at 34.00 suggests an audit of the evidence that supports these answers. Does a dedicated page define the work, explain the process, show an actual sample deliverable, clarify geographic scope and connect claims to credible proof? If a buyer asks how an AVM audit differs from ordinary rank tracking, is the answer written plainly on an authoritative, easily discoverable page? These details help humans make decisions and give assistants material they can summarize with fewer ambiguities.

Authority Intelligence at 41.00, the highest of the six pillars, suggests there is material to build upon. However, the framework does not turn every third-party mention into proof. Earned credibility depends on whether an interview, conference appearance, industry article, client case study or professional citation actually supports the claim being made. Independent validation is more persuasive when it is specific and authentic.

The relationship among the layers matters more than any single score. A structured profile with perfectly consistent company details may still lack service evidence. A rich content library may still be hard to retrieve if important pages are blocked or poorly organized. An excellent case study may not help a category comparison if no relevant service page connects readers to it. VEM is useful because it turns those disconnected website and reputation problems into one coherent entity-readiness conversation.

Advanced VEM: where the entity ecosystem needs reinforcement

The Advanced VEM screen reports an overall score of 35.30/100, labelled Needs Improvement. It then separates the broader problem into nine areas. This lower advanced score should not be mathematically averaged with the base VEM score or treated as a contradictory reading. The panels assess different groups of signals and should be compared within their own defined methodology.

Figure 18. Advanced VEM score summary: an overall score of 35.30/100 and nine component scores covering entity ecosystem, knowledge graph, AI readiness and GEO readiness. Image credit: ThatWare, supplied screenshot.

Advanced VEM indicator	Score /100	Practical meaning for optimization
Entity Ecosystem Analysis	39.00	Assess completeness and agreement among owned pages, authoritative profiles and independent brand references.
Knowledge Graph Strength	38.00	Clarify the relationships among brand, founder, services, research, products and verifiable external identifiers.
AI Search Readiness	36.00	Check technical accessibility, page clarity, internal linking, structured data and answer-ready source material.
Brand Entity Consistency	37.00	Identify contradictory descriptions, outdated profiles, duplicate service names and mismatched company details.
Competitive Entity Gap	32.00	Investigate the brand’s weaker position against selected comparable entities. This is not a measured 32% market deficit.
Query Intent Coverage	32.00	Expand legitimate coverage of discovery, comparison, problem-solving and purchase-stage intent.
Semantic Content Strength	34.00	Improve service definitions, supporting explanations, examples and the links between related topics.
AI Citation Probability	35.00	Treat as a citation-readiness estimate in this framework, not a calibrated 35% likelihood of citation.
GEO Readiness	33.00	Assess whether SEO, answer-focused content, entity clarity and supporting evidence form a coordinated optimization program.

The 32.00 scores for Competitive Entity Gap and Query Intent Coverage point toward the same marketing question: can the brand win consideration beyond branded searches? If generic and comparative queries do not consistently connect ThatWare with its priority services, a technical tweak alone will not bridge that gap. The commercial site architecture, topical evidence, and quality of category-level explanations must work together.

Knowledge Graph Strength at 38.00 and Brand Entity Consistency at 37.00 raise a different question: do the same facts connect correctly across the digital ecosystem? The company name, founder relationship, service families, AVM and VEM frameworks, keynote contributions and published resources should be documented consistently where appropriate. Structured data can make supported relationships easier to parse, but it should agree with visible page content. Google’s own AI feature guidance does not require special AI files or proprietary schema to qualify for its AI Overviews and AI Mode results.

GEO Readiness at 33.00 and AI Search Readiness at 36.00 should therefore be read as a prioritization signal, not a call to create dozens of experimental files. Technical discoverability, strong accessible pages, helpful original content and consistent descriptions remain fundamental. Optional machine-readable assets may be useful for a particular workflow, but they are not a substitute for credible primary-source information and do not guarantee inclusion in an AI response.

For businesses comparing GEO services, AI visibility audit services or entity SEO consulting, this is a valuable distinction. A provider should be able to explain which observed answer is weak, what evidence is missing, why the proposed fix addresses that gap and how the result will be tested again. The number without the diagnosis is not a strategy.

What these three scores say together, and what they do not say

Consider the three headline readings side by side: AVM 35.86, VEM 36.85 and Advanced VEM 35.30. All three point to a developing or weak position under their respective frameworks, but they describe different things. AVM focuses on observed treatment within a sampled AI output environment. VEM studies the supporting entity foundation. Advanced VEM expands the entity analysis into more specific operational gaps.

The most defensible conclusion is not “OpenAI has ranked ThatWare 35th out of 100” or “ThatWare has a 35% chance of being recommended.” Neither statement follows from the dashboard. The defensible conclusion is that within this scoped assessment, basic recognition outpaces citation and comparison strength, while several entity and query-coverage signals need work.

That reading also shows why the phrase AI visibility needs more precision in a client meeting. A company with a strong Presence score may still be weak in Citation. A company that appears in answers may still have low query-intent coverage. A stable brand mention may still be attached to thin or outdated evidence. AVM and VEM are most useful when they explain the disconnect and point to a testable correction.

There is another important limitation. The dashboard is provider-labeled, not a full-market benchmark. An OpenAI-scoped assessment does not automatically establish the company’s status in Google AI Overviews, Gemini, Claude, Perplexity or every regional search environment. Responses can vary by prompt wording, retrieval method, model version, country, language and date. An honest executive report records those variables instead of masking them with one universal-looking score.

How to report AVM and VEM results without misleading clients

The temptation with any 0-100 score is to create a tidy before-and-after story. AI answer measurements demand more discipline. A score is meaningful only within the method, evidence and test context that produced it.

In practical terms, we recommend keeping an audit record that identifies the prompt universe, markets and languages; provider or surface; model version when known; run dates; included competitors; any weighting or thresholds; and the raw answers and citations supporting the final summary. If those inputs change, explain the change before drawing conclusions from movement in the score. A result from a provider API or diagnostic service may not reproduce what every customer sees in a consumer interface.

Some metric names deserve particular care. AI Memory does not establish that a platform permanently remembers a brand. Answer Probability and AI Citation Probability are not calibrated probabilities unless validated as such. AI Market Share Visibility is not financial market share. Confidence is about a report’s evidence and should not be described as customer trust. Those distinctions are not disclaimers to hide in fine print; they determine whether decision-makers can use the report responsibly.

A credible measurement program also keeps traditional business outcomes in view. Qualified organic sessions, lead quality, branded demand, sales conversations, enquiries and conversions still matter. AVM and VEM extend the view into how a brand is represented before a click. They do not replace a commercial performance dashboard or establish that every score change caused revenue movement.

The commercial lesson: visibility is the starting line, not the finish line

The BrightonSEO keynote argued that citation is an output signal while the brand conclusion sits further along the decision process. This dashboard gives that argument a practical form. The AVM report shows basic presence alongside much weaker citation and position. The VEM views point toward gaps in query coverage, content depth, entity connections and GEO readiness. Together, they suggest an action plan rooted in stronger evidence and clearer buyer relevance.

At ThatWare, we use AVM to ask what AI systems are visibly doing with a brand and VEM to investigate whether the brand’s underlying identity and evidence are ready to support better answers. The two approaches are complementary, not competing scores. They help turn a vague request to “rank in ChatGPT” into a specific, accountable program: choose meaningful prompts, observe real outputs, validate the claims, repair the evidence and test again.

If your brand is being mentioned but rarely cited, or cited but seldom shortlisted, an AI search visibility audit can help identify where that transition breaks down. The goal is not to manufacture a perfect score. It is to make the company’s real capabilities easier to find, understand, verify and recommend for the right customer need.

A practical 90-day roadmap for AI brand visibility

Priority from this sample	Primary workstream	Evidence of a useful improvement
Low Citation: 13.33	Claim-specific supporting assets, original research, service proof and clear source architecture	A higher share of relevant sampled answers cites pages that actually substantiate the claim.
Weak Position: 26.67	Commercial comparisons, differentiated service positioning and buyer-fit explanations	More accurate inclusion in appropriate shortlists and better documented comparative standing.
Weak Query Intelligence: 32.00	Mapping service-intent prompts to answerable commercial pages	Broader, accurate coverage of priority non-branded questions in the fixed prompt universe.
Weak AI Trust: 31.00	Verifiable credentials, named methodology, independent context and editorial governance	More supported, accurate assessments rather than generic praise or unverified endorsements.
Weak GEO Readiness: 33.00	SEO, AEO, entity consistency, technical access and evidence coordination	Fewer factual ambiguities, better retrieval eligibility and a consistent retest record.

The most sensible first quarter is not about promising to dominate every model. It is about establishing a baseline, fixing the largest information gaps and testing whether meaningful responses improve.

Days 1-15: establish a defensible baseline

Agree on the most important commercial journeys, markets and buyer personas. Create the prompt universe and select competitors. Sample relevant AI systems with documented methods. Review Search Console and existing search performance. Build a citation map and identify important answer inaccuracies. Capture the current state before significant changes so the team has a credible comparison point.

Deliverable: baseline visibility report, prompt library, factual-error register and prioritized list of high-value query categories.

Days 16-35: repair identity and access

Resolve canonical brand, service, product and location inconsistencies. Audit technical indexing and accessibility. Strengthen internal links to commercial pages. Correct high-impact structured-data mismatches and known external listing errors. Define approved facts and who owns them. Fix public contradictions that could undermine buyer confidence.

Deliverable: canonical entity inventory, technical repair list, approved brand fact sheet and corrected priority pages.

Days 36-60: build the missing evidence

Write or improve the pages that answer actual buying questions. Produce qualification criteria, honest comparison guidance, product specifications, integration details, service methodology content and responsibly sourced examples. Ensure each important claim links to its underlying proof. Build accessible, useful content first; optimize its structure without turning it into unnatural AI-facing copy.

Deliverable: evidence-backed priority pages, improved answer content and claim-to-source map.

Days 61-75: strengthen trustworthy external context

Identify legitimate opportunities for professional coverage, expert commentary, partner documentation, directories, marketplace profiles or other third-party sources appropriate to the industry. Correct factual discrepancies, rather than trying to accumulate low-value mentions. Coordinate with public relations and customer teams to ensure claims and endorsements are permissioned and accurate.

Deliverable: external evidence plan, completed corrections and a maintained outreach and source register.

Days 76-90: retest, compare and govern

Rerun the prompt set with comparable settings, including repeated and paraphrased questions. Review source changes, accuracy, comparison inclusion and qualified recommendations. Document changes in uncertainty as well as positive results. Connect visibility reporting to Search Console and reasonable business indicators. Keep unresolved tasks in the next-quarter backlog.

Deliverable: before-and-after evidence review, updated AVM/VEM diagnostic, executive report and maintenance schedule.

The most important habit is repeatability. At the end of the quarter, leadership should be able to see what was changed, which claims were corrected, what answer behavior was observed, and which uncertainties remain. That is considerably more useful than a promise to “make AI recommend us” without a way to verify it.

Figure 19. A phased 90-day roadmap: baseline, entity repair, evidence creation, external validation and retesting. Credit: ThatWare brightonSEO presentation.

Five mistakes that make AI visibility programs expensive and unconvincing

Mistake one: counting citations as wins regardless of answer quality. A cited page can be used to support a neutral definition while competitors receive all the recommendations. Report source use separately from commercial selection.

Mistake two: promising guaranteed ChatGPT rankings or platform trust scores. Model outputs can change, and outside vendors do not control the providers’ private decision systems. A professional agency should offer transparent methods and measurable hypotheses, not invented access to hidden ranking signals.

Mistake three: publishing a large volume of interchangeable AI content. Content that repeats general knowledge without original value may dilute a site. Google explicitly advises creators to prioritize unique, useful material and warns against scaled manipulation. The strongest pages answer a real buyer problem more clearly than a generic rewrite.

Mistake four: treating experimental files as universal requirements. An llms.txt or custom JSON file can help a particular workflow if a consumer exists, but it should not be sold as a mandatory Google AI Overview optimization tactic. Fix ordinary accessibility, crawlability and factual accuracy first.

Mistake five: ignoring the cost of a wrong recommendation. In finance, healthcare, safety-sensitive products and regulated B2B categories, a flattering but false AI answer can mislead a buyer. Track unsupported claims and unsuitable recommendations with the same attention as visibility gains.

If an SEO or GEO proposal does not define the buyer questions, the evidence requirements, the measurement method and the risk controls, it is not yet a serious AI search strategy.

How to evaluate an AI visibility audit or GEO agency

The right questions are more useful than a generic ranking of agencies. Ask for the audit’s exact scope: which products will be tested, how prompts will be selected, whether branded and unbranded queries are separated, and how often the sample will be repeated. Ask how the provider handles answers with no citations and how it tests whether a cited page supports an important claim.

Next, ask for the deliverables. A useful engagement should produce a clear entity map, prioritized content and technical fixes, evidence requirements, commercial intent coverage and a monitoring framework. It should be possible to hand those tasks to web, editorial, PR and product teams without interpreting a vague technical presentation. Cross-functional execution matters more than proprietary terminology on its own.

Ask how the agency distinguishes results it can observe from outcomes it cannot guarantee. If a supplier claims to control Google’s internal AI scoring, make it explain the source of that claim. Google’s own guidance notes that third-party tools do not have access to Google’s internal ranking or AI systems. A credible AI search optimization agency should be comfortable saying where the evidence ends.

Finally, consider fit. A local practice, a global ecommerce company and a complex B2B software vendor need very different evidence architectures. The strongest project plan reflects the business model, buyer journey, content maturity, budget and risk level rather than selling the same “AI citations package” to everyone.

For businesses that want to examine this systematically, ThatWare offers GEO consulting and implementation, AEO services and complementary AVM/VEM research frameworks. Our preferred starting point is a question with a clear commercial answer: Where are qualified buyers asking for solutions, and what does AI currently conclude about our brand there?

AI Search Visibility Metrics, Brand Risk Controls and Five Executive Takeaways

Five expanded sections and visuals to complement “It’s Not Enough to Be Cited: How AI Systems Decide Which Brands to Recommend.”

Editorial field	Publishing guidance
Document purpose	Publication-ready additional blog sections, designed to be integrated into the existing evergreen ThatWare article.
Recommended placement	Place the metrics and risk sections after the 90-day roadmap; position the five takeaways and closing CTA near the end of the article, before sources and credits.
Primary search theme	AI search visibility metrics and measurement
Commercial search intent	AI visibility audit; GEO consulting; AI recommendation tracking; AI brand reputation monitoring; LLM SEO agency; generative engine optimization services
Visuals	Five keynote visuals, including a date-free crop of the closing slide; each figure carries a source credit.
Publishing location	Normal ThatWare Blog section. Keep the existing blog H1 and slug; these additions use H2/H3 sections.
Editorial position	Written in ThatWare’s first-person brand voice, with practical evaluation criteria and clear measurement limitations.

Integration note: the editorial cover on this page is not intended for publication; the blog copy begins on the following page.

Visual rights: slide imagery supplied for adaptation; review the rights of embedded external graphics before web publication.

The New Metrics for AI Search Visibility: Measure the Brand Conclusion, Not Just the Citation

A conventional SEO report can tell a marketing director which pages rank, how often those pages appear in search, and how many visitors arrive. Those numbers still matter. They do not answer a different question that has become critical to commercial discovery: when a potential buyer asks an AI system to choose between providers, what does the system actually say about our brand?

A company can appear in ten generated answers and still be described as a secondary option. Another company may appear less frequently but dominate the specific buying conversations that generate qualified leads. The difference becomes invisible when the reporting model treats every mention as an equal success. That is why we recommend measuring the whole sequence, from appearance and attribution to accuracy, comparative position and suitability for a genuine customer need.

At ThatWare, we use the term AI visibility metrics for a practical set of observed outcomes. They are not secret ranking signals, official scores issued by ChatGPT or Google, or guarantees that a model will make the same decision next time. They are research and reporting measures that a company can define, audit, repeat and connect to business priorities. This distinction is important when evaluating an AI visibility audit, an LLM SEO agency, or a generative engine optimization service. A dashboard is only as credible as the prompts, observations and definitions behind it.

Figure 20. The six dimensions of an AI visibility scorecard: mention rate, citation rate, positive conclusion rate, recommendation rank, source diversity and entity consistency. Original presentation visual: ThatWare brightonSEO keynote.

1. Mention rate: are we present in the answers that matter?

Mention rate is the share of tested answers that explicitly identify the target brand. The simplest working definition is: answers that mention the brand, divided by all valid answers in the selected prompt set, expressed as a percentage. Before collecting results, decide whether the brand must be named exactly or whether approved name variants and product names qualify. An unrelated use of a common brand word should not count.

For commercial teams, mention rate is most helpful when divided by intent. Separate prompts such as “What is GEO?” from prompts such as “Which GEO agency should a mid-market B2B company evaluate?” The first tests topical exposure. The second tests potential buyer consideration. A rising overall mention rate may look impressive while the high-intent shortlist remains unchanged.

Also segment by engine, location, language and prompt type. A regional services business may be well represented in city-specific questions yet absent from the national comparison prompts its prospects use. A single blended percentage would hide that difference. Mention rate is therefore a useful entry metric, not a conclusion about commercial success.

2. Citation rate: which sources are being used as evidence?

Citation rate asks how often the brand’s owned or designated authoritative sources are linked or explicitly cited in the responses being evaluated. Define the denominator consistently. For example, calculate the share of valid responses containing at least one qualifying citation, then separately record the number of citation occurrences and the unique URLs involved. A response with four links to one domain should not be mistaken for four independent sources.

Citation analysis should extend beyond counting links. Record which exact page is cited, which claim appears near the citation, whether the source truly supports that claim, and whether the citation points to the original publisher, a third-party article or an aggregator. An answer that cites a brand’s guide for a definition is different from one that relies on its service evidence to justify a recommendation.

For websites, it is worth connecting citation observations to first-party search data rather than replacing one with the other. Google’s dedicated generative AI performance reporting provides visibility into appearances of pages within eligible Google AI search features, including impressions and URLs; it does not by itself constitute a comprehensive measure of how every independent assistant describes or recommends a business. Google Search Central: Generative AI performance reports.

3. Positive conclusion rate: what judgment does the answer leave behind?

Positive conclusion rate is the share of answers in which the system makes a favorable, accurate, and contextually appropriate assessment of the brand. We recommend scoring this with a predefined rubric rather than relying on a generic sentiment model. “A recognized provider” is not necessarily a recommendation. “A suitable option for regulated enterprise deployments because of documented controls” is much closer to a decision-relevant positive conclusion, assuming that the evidence really exists.

A usable classification scheme might label each response as positive, neutral, mixed, negative, or not assessable. The reviewer should capture the actual sentence that supports the label. Positive conclusions must also pass a factual accuracy check. An answer that recommends a brand for a capability it does not offer is not a marketing victory. It is a governance concern.

The metric is most useful when paired with the prompt’s job to be done. For one prompt, a favorable result is inclusion in a reasonable shortlist. For another, it is accurate recognition of a specialist certification. For a high-risk service, the best outcome might be a cautious, qualified explanation rather than enthusiastic promotion.

4. Recommendation rank: where do we sit when AI compares alternatives?

Recommendation rank records the brand’s position within ordered lists or clearly comparative responses. It should be calculated only when the answer actually ranks or sequences options. If an AI tool lists five vendors alphabetically or names them without preference, assigning an artificial rank would distort the evidence. Store both the observed position and the reason the brand was included or excluded.

Teams can report top-one inclusion, top-three inclusion, average observed position in ordered lists, and exclusion rate. These numbers must be segmented by query intent. Being first in a broad educational answer does not carry the same commercial value as being named among the leading candidates for a purchase-ready request.

A practical review question is: “What evidence would make our brand a defensible choice for this particular user?” If competitors consistently win for implementation support, service coverage or third-party validation, the remedy may be better documented capabilities, not more keyword repetition. Recommendation rank converts an otherwise vague visibility complaint into a specific competitive research question.

5. Source diversity: do different kinds of credible evidence reinforce the brand?

Source diversity measures how many relevant categories of credible evidence support the brand, rather than how many URLs repeat the same statement. An owned service page, independent trade publication, verified customer review, product documentation and professional certification are different evidence classes. Ten syndicated copies of one press release are not ten independent validations.

To report diversity responsibly, define your evidence taxonomy and record a source as present only when it is genuinely relevant, accessible and supportable. A source can be accurate without being independent; conversely, a third-party post can be independent without being reliable. Quality and independence must be assessed separately. We suggest tracking both the number of evidence classes represented and the proportion of material claims that have corroboration outside owned assets.

This metric matters most for claims that buyers are unlikely to accept at face value, including comparative performance, coverage claims, technical standards, pricing promises and case-study outcomes. Where independent evidence is unavailable, the correct editorial action is to narrow the claim or publish verifiable first-party proof, not to create artificial endorsements.

6. Entity consistency: does the same business emerge across engines and prompts?

Entity consistency captures whether AI responses accurately connect the brand to the correct services, locations, products, executives, ownership and other stable facts. It is a stronger diagnostic than sentiment alone. A model may sound positive while confusing a company with a similarly named competitor, attributing a retired service to its current offering, or naming the wrong headquarters.

Prepare a compact, approved fact sheet for scoring: canonical brand name, primary site, core services, geographic coverage, current product portfolio, supported credentials, and a last-reviewed date. Rate each extracted claim as correct, incomplete, contradictory, or unverifiable. For each error, save the prompt, answer text, model or surface, locale and cited URL if available.

Entity consistency is a natural bridge to ThatWare’s Vector Entity Modelling (VEM) approach, which focuses on how clearly a brand and its relationships are understood across digital evidence. A page-level fix can correct one statement. An entity-level review looks for the conflict across site architecture, profiles, structured information and third-party references. For more context, see ThatWare Labs: AVM, VEM and GEO.

Make the metrics comparable: a measurement design that can survive an executive meeting

A useful scorecard begins before the first prompt is run. Select a stable prompt universe and label the intent of each question: informational, comparison, commercial, local, problem-led or brand-led. Add the customer segment and the intended buying stage. For example, a cybersecurity vendor might test “best SOC 2-ready solutions for a 100-person fintech” separately from “what is SOC 2?” because those queries represent very different choices.

Next, control the testing conditions as far as the platform allows. Capture the engine or product surface, timestamp, geography, language, available personalization setting, and whether the result was generated from a fresh conversation. Repeat representative prompts over time because models and retrieval results are variable. A one-time screenshot is an observation, not a stable market share estimate. For higher-stakes conclusions, use a clear sample size, multiple runs and human validation of classifications.

Finally, connect the observations to commercial outcomes without inventing attribution. A recommendation seen in a generated answer does not prove a sale resulted from that answer. Use it alongside branded-search trends, lead quality, influenced pipeline, customer interviews, direct traffic patterns and existing SEO analytics. A strong report states what was observed, what is inferred and what remains unknown.

Metric	Suggested reporting unit	Business interpretation
Mention rate	Percentage of valid prompt responses naming the brand	Basic share of answer presence
Citation rate	Percentage containing a qualifying cited source, plus cited URLs	Visibility of supporting evidence
Positive conclusion rate	Percentage of scored responses with a supported favorable judgment	Quality of brand portrayal
Recommendation rank	Top-one/top-three frequency in genuinely ordered answers	Shortlist competitiveness
Source diversity	Number and quality of independent evidence classes	Breadth of corroboration
Entity consistency	Percentage of checked material facts stated accurately	Reliability of brand understanding

A practical example: why two brands with the same citation rate can have very different outcomes

Consider a hypothetical group of 100 commercial-intent prompts about enterprise project management tools. Brand A appears in 42 responses and receives source citations in 27. Brand B appears in 31 and receives citations in 26. On those two numbers alone, Brand A looks stronger. But imagine Brand A is recommended in only 9 answers, while Brand B is recommended in 19, and that Brand B’s answers more consistently explain product fit and implementation strengths.

The sample is illustrative, not a result from a live ThatWare client study. Its purpose is to expose the reporting trap: a citation-heavy strategy can make the organization more visible without making it more convincing. A business should use the missing recommendation evidence to decide whether to improve service specifications, case studies, comparison pages, external validation or the factual consistency of its brand footprint.

Figure 21. From traditional SEO rankings to retrieval, synthesis and AI recommendation visibility. This diagram illustrates how webpage content, documents, structured data, authority signals and entity relationships can contribute to an answer. Credit: ThatWare brightonSEO keynote presentation.

Reading the journey from rankings to AI recommendation visibility

The second visual shows why the reporting system needs several layers. On the left is a familiar search-results view, where marketers concentrate on page order and clicks. The center represents a simplified view of retrieval, synthesis and comparison. On the right is a generated answer assembled with supporting links and brand-specific judgments. Real services do not all use the same architecture, and the diagram should not be read as a disclosure of any provider’s proprietary model. It is a teaching model for the different outcomes we must observe.

The practical lesson is that webpage authority, structured facts, third-party evidence and the user’s problem can influence an answer in different ways. An AI system might retrieve a brand’s documentation, use a competitor’s independent review for comparison, then recommend a third provider whose documented constraints better match the user. More visibility at one stage does not automatically guarantee selection at the next.

This also reinforces the role of foundational SEO. Google explicitly advises that crawlability, index eligibility, useful content, internal links and a good page experience remain relevant to its generative search features. Google does not require special AI text files or an AI-only schema to appear in those features. A sensible GEO strategy therefore starts with trustworthy, accessible information and adds disciplined measurement of AI outputs rather than promising a shortcut around core search fundamentals. Google Search Central: Generative AI optimization guide.

For a marketing leader commissioning AI search visibility tracking, the deliverable should not be a single “visibility score” with unexplained weights. Ask for the raw prompt sample, answer captures, scoring definitions, cited URLs, evidence review, segment-level trends and a clear statement of limitations. When the team can show its work, the metric becomes something management can act on.

Risks Brands Must Control Before AI Controls the Narrative

What happens when the publicly available evidence about a business is inconsistent? Sometimes nothing dramatic occurs: the assistant gives a vague answer or omits a provider. In other cases it may repeat an outdated promise, mix together two businesses, or present a claim more confidently than the source deserves. The risk is not simply that an organization loses a citation. It is that a potential customer receives a misleading explanation at the exact moment they are deciding whom to trust.

Most of these problems cannot be solved by instructing a third-party AI system to “prefer” the brand. They require the company to improve the information available for verification, identify the source of contradictions and monitor how the public-facing narrative changes. That work sits at the intersection of content strategy, technical SEO, review management, digital PR, product marketing and factual governance.

Figure 22. Five risks that can weaken or distort AI-generated brand conclusions: entity confusion, unsupported claims, third-party drift, citation hijack and freshness decay. Credit: ThatWare brightonSEO keynote presentation.

Risk 1: Entity confusion can put the wrong business in the answer

Entity confusion occurs when information belonging to one organization, person or product becomes mixed with another. It is especially common as a practical risk for brands that share names, have legacy domains, operate under subsidiaries, or use inconsistent product labels. An assistant might describe a regional consultancy using the service portfolio of a company with a similar name. Even a flattering answer becomes commercially useless when it refers to the wrong organization.

Start by maintaining a current canonical identity record and comparing it with priority website pages, business profiles, professional directory entries, relevant knowledge sources and company-owned social accounts. Keep official service names, business relationships and location information consistent. Where appropriate, use conventional structured data that accurately matches the visible page. Do not assume a sameAs link or schema field can force a model to resolve identity correctly. It can provide clearer machine-readable context, not a guaranteed result.

For an AI visibility audit, measure both false associations and missing associations. If the brand is consistently linked to an old service, trace the wording back to the source pages that still make that claim. Fix the underlying content and request updates where the conflicting material is under third-party control. This is a stronger response than publishing another generic “About us” article.

Risk 2: Unsupported claims can turn an apparently positive answer into a liability

A brand may promise the “fastest deployment,” “highest success rate” or “most secure platform” without a published methodology, comparison set or independent proof. If AI repeats that claim, the brand may enjoy a temporary visibility benefit while exposing buyers to unverified information. If AI refuses to repeat it, the marketing team may misdiagnose the problem as an optimization failure when it is actually an evidence failure.

Build a claims register for the statements that matter most to purchase decisions. Each row should identify the claim, its approved wording, supporting source, owner, last verification date and where it appears. Separate factual specifications from opinions and marketing language. If a statement is comparative, explain the basis of comparison. If the proof is confidential or incomplete, narrow the public claim to what can be defended.

The standard is simple: a reader should be able to understand why the business is making the claim. In regulated, healthcare, financial or safety-sensitive categories, legal and subject-matter review is particularly important. The purpose of GEO is not to convince AI to amplify unproved assertions. It is to make verified strengths easier to discover and communicate accurately.

Risk 3: Third-party drift allows yesterday’s reputation to overpower today’s facts

A company can update its website in one afternoon and still be represented by an outdated profile, review, forum discussion or industry article months later. That gap is third-party drift. It does not mean every external criticism should be suppressed. It means the brand’s current facts and its wider evidence footprint may be telling different stories.

For example, an IT services firm may discontinue a service and revise its navigation, while older directories still list the retired offer. A prospective customer asks which companies provide that service and receives a confident but inaccurate inclusion. Alternatively, an independent review may highlight a limitation the company has now resolved, while its documentation provides no dated explanation of the change.

Maintain an external source inventory ranked by relevance, credibility and risk. Correct profiles you own, approach publishers with well-supported updates when appropriate, and publish accessible change logs, updated specifications and current case studies where they help customers. Do not fabricate reviews, coordinate inauthentic commentary or attempt to bury legitimate negative feedback. Authentic public evidence is more durable than reputation manipulation.

Risk 4: Citation hijack can make aggregators the apparent authority on your category

Citation hijack is our shorthand for a commercial situation, not a recognized technical attack pattern: an intermediary, aggregator or competitor becomes the repeatedly cited source for information that the brand could have explained directly and more accurately. This can happen when the organization’s product pages are thin, its comparison details are missing, or external sites provide clearer answers to buyer questions.

Imagine a manufacturer with detailed PDF specifications but weak HTML explanations of compatibility and installation constraints. A buyer asks an AI assistant to compare product options. The assistant may lean on a distributor’s summary because that is the easier accessible source for the question. The manufacturer’s facts are still involved in the market conversation, but its own voice is less visible and its product differences may be flattened.

The remedy is not to block every intermediary or demand exclusive citations. Give important commercial topics a strong canonical home: a useful comparison page, readable technical specifications, transparent FAQs, supporting diagrams, case studies and clear links between product models and applications. Then assess whether outside summaries are accurate. Where appropriate, work with legitimate industry partners to improve their information too.

Risk 5: Freshness decay lets historical facts answer present-day questions

Freshness decay becomes a problem when prices, policies, product coverage, team credentials, availability or capabilities change but the public evidence remains frozen. A model can only work with the information it retrieves or has learned through its system’s own processes. It may encounter both new and old pages, or no current evidence at all. Neither an updated timestamp nor a self-declared “last reviewed” note can compensate for incorrect substance.

Create a maintenance priority list based on the cost of an outdated answer. Product specifications and eligibility conditions may need a tighter review cycle than an evergreen glossary. Assign a factual owner to each high-impact page, keep publication and revision dates truthful, and retire contradictory legacy content using appropriate redirects or clear archive labeling. Update internal links and structured information in step with the visible copy.

The core governance question is not “How often should we refresh content?” It is “How quickly would we discover and correct an answer that is no longer true?” That turns maintenance into a risk-management process rather than a calendar-driven rewrite exercise.

The control system: monitor, verify, correct and retest

A brand risk register should be tied to the same prompt library used for visibility reporting. When an incorrect answer is found, preserve the exact prompt, full answer, evidence links, surface, locale and observation time. Mark whether the issue is a factual error, misleading omission, unsupported endorsement or confused entity. Then investigate the available public source material before selecting a remedy.

Control	What the team should document	Accountable group
Detect	Prompt, answer, citations, location and observed impact	Search/AI visibility analyst
Verify	Correct fact, dated proof and severity of error	Product or subject-matter owner
Correct	Canonical page, profile, document or publisher update	Content, SEO and communications
Retest	Repeat comparable prompts and track changes over time	Analytics and quality assurance
Escalate	Legal, safety or material customer impact	Leadership and compliance, if relevant

Use severity rather than raw error count to prioritize work. Confusing a business’s former office address with its current office address is inconvenient. Recommending an unqualified provider for a safety-critical service can be materially worse. Define escalation thresholds, review ownership and a retest schedule before a problem occurs.

No company can fully control third-party models or guarantee that corrections will appear immediately. But a disciplined AI brand reputation monitoring process can shorten the time between discovering a problem and repairing the evidence that contributed to it. It also gives teams a defensible record of what they observed and what they changed.

Five Takeaways for Brands: What Our BrightonSEO Keynote Means in Practice

Our BrightonSEO keynote condensed a much wider commercial challenge into five ideas. Each one changes a decision that marketing, SEO or executive teams make. Taken together, they suggest a more useful way to invest in organic visibility: improve the clarity and credibility of the brand’s public evidence, then measure whether that evidence is supporting appropriate inclusion in real buyer conversations.

Figure 23. Five core takeaways: citation is not recommendation; AI sources differ from organic rankings; brand conclusions depend on verifiable evidence; SEO, AEO and GEO must connect; and AVM/VEM measurement must be repeated. Credit: ThatWare brightonSEO keynote presentation.

Takeaway 1: being cited is not the same as being trusted or recommended

A mention shows presence. A citation shows a source relationship. Neither proves that an AI assistant thinks a brand is suitable for a particular customer. If your executive dashboard shows a rise in citations but never examines the wording of comparative answers, it may celebrate visibility without noticing lost consideration.

For every priority commercial query, classify the observed outcome: absent, mentioned, cited, described, compared or recommended. Some answers will occupy more than one category. Preserve the wording that supports the classification and verify whether the recommendation is accurate. A good result is not simply a positive adjective. It is an evidence-backed explanation that connects a genuine strength to a genuine customer need.

Action for leadership: ask the next AI performance report to show the five most consequential recommendation wins and losses, with the prompts and supporting sources attached. That one change makes a conversation about GEO substantially more useful.

Takeaway 2: AI systems may select sources differently from traditional ranking lists

High organic visibility is valuable, but it should not be treated as an automatic proxy for AI answer inclusion. Search providers can use different retrieval approaches, and an answer assembled for a nuanced question may draw from a different set of pages than a familiar ten-result ranking. Google’s documentation describes query fan-out and retrieval-augmented generation for its generative experiences, along with the continuing importance of its core search systems. Google Search Central: Optimizing for generative AI.

This means the SEO team should still protect crawlability, index eligibility, helpful content and internal linking. The additional work is to inspect which sources appear for the questions customers really ask and identify whether the brand’s own information provides the evidence needed to answer them. Technical SEO and GEO should inform each other, not compete for budget as unrelated departments.

Action for leadership: require a source-level gap analysis for a handful of commercially valuable prompts. Determine whether missing visibility is caused by technical access, weak content, insufficient corroboration or a genuine lack of product fit.

Takeaway 3: brand conclusions depend on clear identity and credible, consistent evidence

A business can manufacture many pages about itself. It cannot manufacture independent proof that all its promises are true. An AI-facing brand strategy should therefore concentrate on the strength and agreement of its evidence: clear service definitions, original case studies, current specifications, transparent methodology, credible third-party coverage and a coherent public identity.

The case for better evidence is not purely technical. A human procurement officer also wants to know which solution fits, what it costs to adopt, what limitations exist and why the provider is trustworthy. Better commercial content helps both the visitor who lands on the website and the system that draws on publicly accessible information to generate a response.

Action for leadership: choose the ten highest-value buying claims and insist that each has a named owner, a current proof source and a sensible public explanation. Remove or qualify claims that cannot be substantiated.

Takeaway 4: SEO, AEO and GEO must share one knowledge strategy

SEO remains responsible for discoverability, accessibility, information architecture and search performance. AEO emphasizes useful, direct responses to user questions. GEO examines how a brand is represented when systems synthesize information and present comparisons, citations and recommendations. Different labels can describe overlapping work, particularly within Google Search; they should not become three disconnected content calendars.

We recommend organizing these disciplines around the same product and customer truth. A commercial service page should clearly say what the company does. Supporting content should answer practical buying questions. Case studies and documents should demonstrate results and limitations. Structured data should accurately reflect visible content. External profiles should not contradict the website. A shared governance process makes the whole system more reliable.

Action for leadership: use one prioritized backlog, with each task mapped to a customer question, a canonical page or evidence asset, the responsible owner and an observable success criterion. This reduces duplicate work and makes agency accountability clearer.

Takeaway 5: measure AVM and VEM repeatedly because AI answers are dynamic

The language in a generated answer can change with the prompt, the context, the available evidence, the service and the time of testing. That is why a snapshot should never be presented as a fixed market position. The correct comparison is repeated measurement using disclosed methods and suitably controlled conditions.

Within ThatWare’s research framework, AVM (AI Visibility Metric) focuses on how a brand appears, is cited and is discussed across monitored answer environments. VEM (Vector Entity Modelling) focuses on the clarity and consistency of the relationships through which a brand can be understood. These are ThatWare frameworks for analysis, not official scores published by search or LLM vendors. See ThatWare Labs for the company’s current description.

Action for leadership: establish a baseline, make targeted evidence and entity changes, then retest the same high-value questions. Include uncertainty and samples in reporting. A recommendation-rate improvement across a reliable panel is more meaningful than an unexplained jump in a proprietary “AI score.”

Where to put budget first

If the company has strong rankings but weak AI recommendations, we would not automatically start with another volume content campaign. We would first examine whether its commercial claims are easy to verify, whether important comparisons are addressed honestly, and whether relevant third-party sources tell a consistent story. If the company is frequently recommended for services it does not offer, the priority shifts to factual correction. If it is absent because key pages cannot be indexed, technical access takes precedence.

A simple budget triage helps avoid spending money on the wrong symptom. First protect the ability to be discovered. Next make the organization’s core facts coherent. Then produce original evidence for decision-stage questions. After that, strengthen legitimate external corroboration and monitor outcomes. This order will vary by business, but it is better than buying a fixed package of “AI citations” without knowing whether citations are the problem.

The five takeaways come back to the same principle: a brand earns lasting relevance by giving customers and information systems sound reasons to understand it, compare it fairly and, when appropriate, choose it. The job of an AI visibility strategy is to make those reasons findable, accurate and measurable.

From the BrightonSEO Stage to a Practical AI Visibility Audit

We presented a shorter version of these ideas during Dr. Tuhin Banik’s brightonSEO keynote. The point of the presentation was not that rankings no longer matter, or that a citation has no value. It was that modern brand discovery needs a more complete performance question: What is the system helping the buyer conclude about us, and can that conclusion be justified by evidence?

For businesses evaluating GEO consulting, AI visibility audit services, answer engine optimization or LLM search optimization, ThatWare’s suggested starting point is a structured assessment of the answers customers encounter. We can examine the commercially important prompt universe, map the brand’s sources and factual claims, assess recommendation outcomes, identify entity inconsistencies and prioritize the changes most likely to improve decision-stage clarity.

This work should remain grounded in the product and service reality of the business. No ethical audit can promise selection by every model, predict all future answers or turn unverified marketing assertions into facts. What it can deliver is a transparent baseline, a sharper understanding of the gaps, an evidence-backed improvement plan and a sensible way to measure progress over time.

If you want to know whether AI assistants are merely citing your company or actually giving customers reasons to consider it, explore ThatWare’s GEO methodology, learn about our AVM and VEM research, or contact ThatWare. You can reach the team at info@thatware.co. The goal is not to manufacture a preferred answer. It is to make the facts about your brand easier to retrieve, verify and represent correctly.

The brand that wins is the brand that can be understood and justified

A company cannot purchase an enduring place in every AI-generated recommendation. It can, however, make its identity clearer, document its strengths honestly, expose credible evidence, correct contradictions and answer the exact questions buyers ask before making a decision. Those are worthwhile investments even when the answer engine changes tomorrow.

Our central point is deliberately sharper than “get more citations.” The commercial goal is to earn accurate consideration and appropriate selection. Citation data is useful evidence about one part of that process. It should not become the entire strategy or the entire definition of success.

We introduced this idea in a shorter keynote at brightonSEO because we believe the industry needs a more disciplined conversation about AI-driven brand discovery. The expanded framework is a practical starting point: measure retrieval and citation, inspect claim support, map the brand entity, build missing proof and retest the conclusion. When SEO, AEO, GEO and responsible measurement reinforce one another, brands can compete on clarity and verifiable substance rather than hopeful prompt engineering.

If your company is being quoted but not shortlisted, the question worth investigating is not just “How do we get cited more often?” It is “What would a buyer, or an answer system working on that buyer’s behalf, still need to verify before recommending us?”

For a structured assessment of those gaps, explore ThatWare’s GEO services or contact our team to discuss an AI visibility and brand conclusion audit.