AI visibility is not a switch—and a mention is not a customer
A founder asks Google, ChatGPT or another answer engine for suitable providers and sees competitors instead. The common response is to buy a “GEO package”, add more schema or publish dozens of near-identical question pages.
That skips the real decision. Search and AI systems first need access to the page. They then need enough clarity and evidence to understand when the business is relevant. Even if the business is mentioned, the description may be wrong, the cited page may be unhelpful or the visitor may never reach a credible commercial next step.
Google says established SEO practices still apply to AI Overviews and AI Mode. A supporting page must be indexed and eligible for a search snippet; there is no special AI schema, machine-readable file or extra technical requirement. Google also says eligibility never guarantees crawling, indexing or appearance. Google Search Central: AI features and your website.
OpenAI says ChatGPT search placement is not guaranteed. A site can be eligible when OAI-SearchBot is allowed and the host or content-delivery network permits traffic from OpenAI's published searchbot addresses. OpenAI: searching the web with ChatGPT.
My verdict: treat discoverability as a commercial evidence chain, not a collection of AI tactics. The objective is to help a suitable buyer find, understand, verify and act—not merely to produce a screenshot of a brand mention.
The Discoverability-to-Demand Chain
I use six gates to locate the earliest reason a credible business remains invisible or commercially irrelevant. Work in order. Later gates cannot repair a blocked, vague or untrustworthy foundation.
1. Access creates eligibility, not visibility
Check that priority service and resource pages return successfully, have one canonical address, appear in the sitemap and remain available in readable text. Important pages should be reachable through internal links, not only a sitemap. Confirm that robots rules, the host and any security layer allow the crawlers the business intends to serve.
Do not confuse Googlebot, OAI-SearchBot and model-training controls. They have different purposes. The practical owner question is simpler: which discovery surfaces matter to customers, and have we deliberately allowed or blocked the relevant access? The guide to deciding whether to block AI crawlers explains that governance choice.
2. Clear business facts make accurate representation possible
The homepage, About page, service pages, author profile and reliable business listings should agree on the customer, problem, offer, locations served and accountable person. Structured data can describe those visible facts, but it cannot convert vague positioning into clarity or unsupported claims into authority.
3. One page should own one buyer decision
Useful pages answer decisions such as how much to invest, which channel fits, why demand is weak or how to select a provider. Lead with a direct answer, explain conditions and trade-offs, and provide a framework the reader can use. The query may be longer and more conversational in AI search, but the page still needs a distinct job.
Google says AI features may issue multiple related searches across subtopics and data sources. That makes complete decision coverage useful, but it is not permission to manufacture every wording as a separate URL. Consolidate overlap and keep the best page internally connected to the wider topic.
4–5. Owned evidence and independent corroboration do different jobs
Owned evidence includes documented cases, transparent methods, named expertise, limitations and current business facts. Corroboration comes from accurate third-party profiles, customer reviews, partnerships, coverage or citations. Repeating the same claim across the company's own pages is not independent support.
Use the guide to customer reviews and search visibility when reputation evidence is weak, and the separate guide on earning business mentions in AI search when access and page quality already pass.
6. Discovery only matters when the right person can act
A cited article can build trust without producing an immediate click. A visit can also be commercially useless when the service, proof and next step do not match the question. Connect high-intent pages naturally to the relevant service, a documented case, the operator behind the work and a stage-appropriate conversation.
Keep three related search jobs separate
This page owns the foundation: whether a business can be found, understood and connected to a useful customer decision. Two adjacent pages answer different questions.
| Owner question | Primary job | Use this resource |
|---|---|---|
| Can search and AI systems find and understand us? | Access, entity clarity, decision coverage and commercial path | This Discoverability-to-Demand Chain |
| Why are competitors mentioned or cited instead? | Answer usefulness, proof and third-party corroboration | AI Citation Readiness System |
| How do we measure whether AI visibility is improving? | Repeated monitoring, citation accuracy, referrals and commercial outcomes | AI Visibility Evidence Ladder |
The investment case is separate again. Use whether SEO still matters with AI search when the decision is how organic discovery should fit the wider growth portfolio.
Diagnose the first weak gate before adding content
| Business symptom | Likely first weak gate | Best next evidence | Avoid |
|---|---|---|---|
| Priority pages are absent from Google | Access | Robots, status, canonical, sitemap, internal links and Search Console | Publishing more pages before repair |
| The company appears but is described incorrectly | Entity clarity | Align the visible facts across core pages and reliable profiles | Adding markup that contradicts the page |
| Competitors surface for valuable buyer questions | Decision fit or evidence | Build one definitive answer with original judgement and verifiable proof | Copying competitor headings into scaled pages |
| The site ranks but is rarely cited in AI answers | Corroboration or retrieval fit | Review authorship, evidence, external accuracy and cited-source gaps | Buying unverified mentions |
| Mentions rise without useful traffic or enquiries | Commercial path | Inspect cited URL, buyer fit, service connection and downstream outcome | Reporting mention count as revenue impact |
If attention already exists but revenue does not, move to the Traffic-to-Revenue Diagnostic. The problem may be offer fit, proof, conversion or measurement rather than discoverability.
Use an evidence ladder, not one AI answer
A July 2026 preprint reviewing 45 studies described generative-engine visibility as a variable, partially observable process and found no reviewed technique that established a stable, longitudinal, cross-platform causal effect on organic discoverability or downstream behaviour. The paper is a critical survey, not a platform rule, but it supports a cautious owner conclusion: one prompt screenshot is weak evidence. Martinez, 2026 critical GEO survey.
Index coverage, crawl access and technical controls are healthy.
Offer, audience, location and evidence remain materially correct.
Repeated buyer questions connect to relevant decision content.
Visits and assisted journeys progress to qualified conversations and revenue.
Google currently includes traffic from AI Overviews and AI Mode within the Web search type in Search Console, rather than providing a separate AI-feature report. Reconcile Search Console with analytics and conversion tracking, then use a small, stable set of buyer questions for repeated AI-answer checks. Record the prompt, market, date, mention, cited URL and accuracy. Compare patterns, not anecdotes.
For the board-level business view, connect this ladder to the Marketing Proof Stack. Discovery is an input. Qualified demand, customer economics and revenue remain the decision.
A practical 30-day discoverability plan
Confirm access
Check priority URLs, crawler rules, hosting controls, sitemap, canonical tags, text visibility and internal links.
Clarify the entity
Align the company, specialist, audience, offer, locations and proof across the core site and dependable profiles.
Strengthen decisions
Improve the three pages closest to revenue with direct answers, original frameworks, evidence and a clear next step.
Start the evidence loop
Connect Search Console, analytics, CRM outcomes and repeated AI-answer checks to one monthly decision report.
Review practical AI growth support, documented case studies, evidence standards and Thomas's direct operating model before choosing a provider or expanding the programme.
Practitioner note: I would not create a new AI-search page until an existing page has a distinct customer decision, current evidence and a commercial job it does not share with another URL. Consolidation is often the more credible optimisation.
Sources and evidence notes
Primary guidance and research were checked on 11 October 2026. The Discoverability-to-Demand Chain, scope map, decision matrix and evidence ladder are original ThomPerformance analysis. Search priority is qualitative; no search volume, ranking guarantee, visibility uplift or client result is claimed.
Frequently asked questions
How can my business appear in AI search results?
Make priority pages publicly crawlable, explain the customers and problems you serve in plain language, support material claims with evidence and keep core business facts consistent across reputable sources. For ChatGPT search, allow OAI-SearchBot. Eligibility helps discovery, but no platform guarantees a mention, citation or position.
Is GEO different from SEO?
GEO, or generative engine optimisation, focuses on how content and businesses appear in AI-generated answers. The dependable foundations overlap with sound SEO: crawl access, clear page intent, useful answers, credible authorship, internal links and evidence. Google says its AI features do not require special optimisation beyond established SEO practices.
Do I need special AI schema or an llms.txt file?
Not to appear in Google AI Overviews or AI Mode. Google says there is no special schema, AI text file or machine-readable markup requirement. Structured data should still match visible facts, and an llms.txt file may document a site, but neither replaces crawlability, internal links, strong pages or independent evidence.
Should I create hundreds of pages for AI search questions?
No. Build one useful page for each distinct customer decision. Strengthen it with a direct answer, first-hand judgement, supporting evidence, clear authorship and natural internal links. Create another page only when it serves a meaningfully different search intent; otherwise consolidate to avoid overlap and weak scaled content.
How should I measure visibility in Google and AI search?
Separate eligibility, accurate representation, mentions or citations, visits and qualified business outcomes. Use Search Console and analytics for Google and referral evidence, plus repeated controlled prompts for directional AI monitoring. Record the cited page and description accuracy, then connect visits to suitable enquiries rather than treating one answer as a stable rank.
How long does it take to become discoverable?
There is no guaranteed timeline. Technical access can be repaired quickly, but recrawling, indexing, authority and independent corroboration may take much longer. Review progress monthly against priority buyer questions and commercial outcomes. Do not refresh dates, manufacture mentions or expand content merely to create the appearance of activity.
Be eligible, understandable and useful before chasing mentions
Google and AI search cannot reliably represent a business that is blocked, vague, unsupported or disconnected from a customer decision. Build the chain in order: access, entity clarity, decision fit, owned evidence, corroboration and a commercial path. Then measure whether accurate discovery creates suitable demand.
Which gate is the first one your current search report cannot prove?