Get you business listed in AI like Chatgpt, Claude & voice Siri
Get you business listed in AI like Chatgpt, Claude & voice Siri
AI search visibility · updated 15 August 2026 How to get your business found in ChatGPT, Google AI & Claude A practical guide to improving visibility across AI-assisted search without relying on made-up AEO shortcuts. The strategy is built around search eligibility, crawl access, useful commercial content, original evidence, clear business information and measurable outcomes. Evidence-led, not theory-led Google AI Overviews + AI Mode ChatGPT Search + Claude Search Technical tests included The most important point: AI visibility is not a replacement for SEO. Google now says this explicitly. The advantage comes from combining strong search fundamentals with content worth citing, clear entity information, accessible crawling and measurement across AI surfaces. Start with evidence Three numbers that change how you should think about AI search A lot of AI-search advice online is recycled guesswork. These figures come from large datasets or primary research, and they point to a much more useful strategy. 37.1% of AI Overview citations also ranked in Google’s organic top 10 Ahrefs analysed roughly 4 million AI Overview URLs across 863,000 keyword SERPs in its 2026 update. Another 26.2% ranked positions 11–100 and 36.7% did not rank in the top 100 for the original query. Independent study · Ahrefs · 2026 8% vs 15% traditional-result click rate when an AI summary appeared vs when it did not Pew analysed 68,879 Google searches from 900 US adults. On searches with an AI summary, users clicked a standard result in 8% of visits; without one, 15% did. User behaviour · Pew Research Center 1% of visits with an AI summary resulted in a click on a cited source The same Pew research found source-link clicks inside AI summaries were rare. This matters because “being cited” and “getting traffic” are not the same commercial outcome. User behaviour · Pew Research Center What this means AI-assisted search changes the value of visibility. Businesses still need strong traditional search signals to be discoverable, but they also need enough topical depth to surface across related retrieval queries, enough trust to be selected as a source and enough brand clarity to remain memorable even when the generated answer reduces the need for a click. What Google actually says AEO and GEO are not magic ranking systems sitting beside SEO Google’s current documentation says its generative search features are rooted in the same core Search ranking and quality systems used by normal Google Search. Google describes retrieval-augmented generation and query fan-out : the system can issue related searches, retrieve useful pages and use them to ground a generated answer. That explains why a page can be cited even if it is not top 10 for the exact prompt. It may rank or be relevant for one of the related searches generated behind the scenes. Claim What the evidence says Recommended action “You must rank top 10 to be cited.” Wrong as a rule. Ahrefs’ 2026 study found 37.1% of cited URLs in the organic top 10 for the original query, while 62.9% were outside it. Build traditional rankings, but also cover the subtopics and evidence likely to appear in fan-out searches. “You need special AEO markup.” Google says there is no special schema or machine-readable AI markup required for generative Search. Use normal structured data where it genuinely describes visible content and supports ordinary Search features. “llms.txt boosts Google AI visibility.” Google explicitly says Google Search ignores llms.txt and that it neither helps nor harms Search visibility. Do not sell or buy llms.txt as a Google ranking tactic. Maintain it only if another system you care about actually uses it. “Write every answer in 150-word chunks.” Google explicitly says there is no chunking requirement and no ideal page length for AI search. Make answers clear and scannable for people. Do not mutilate useful writing to satisfy a made-up word count. Practical conclusion An “AEO package” built mainly around FAQ schema, llms.txt and AI-written Q&A blocks is not a substitute for strong SEO. The harder work still matters: create content that improves on the existing results, keep the site technically accessible, describe the business consistently across the web and publish evidence strong enough to earn trust. Crawler reality ChatGPT Search, Claude and Google do not use the same crawler controls This is where many “AI SEO” guides become technically wrong. Training crawlers and search/retrieval crawlers are not necessarily the same thing. OpenAI OAI-SearchBot is the important crawler for inclusion in ChatGPT Search summaries and snippets. OpenAI says there is no guaranteed top placement, but you should not block OAI-SearchBot if you want your public content to be discoverable and cited. GPTBot is a separate control associated with model improvement/training. Allowing search does not require you to make the same choice for training. Anthropic Anthropic documents three distinct robots: ClaudeBot for model-development crawling, Claude-SearchBot for search, and Claude-User for user-directed retrieval. If you want Claude to retrieve or surface your pages in search-style use cases, blocking the search/user agents can work directly against that goal. Google is different: Google’s generative Search features are built from the Google Search index. The critical crawler is still Googlebot and the page must be indexed and eligible to show a snippet. Google-Extended is not a special “rank me in AI Overviews” switch. Myth-busting Five AI-search tactics to stop treating as shortcuts These tactics are attractive because they are easy to package and sell. Easy does not mean effective. 01 · llms.txt theatre Do not treat llms.txt as a Google ranking factor. Google’s 2026 generative-search guidance explicitly says it ignores the file. If you maintain one for another service, fine; just do not confuse “machine readable” with “ranking signal”. 02 · FAQ schema spam Do not add FAQ schema everywhere expecting extra Google visibility. Google restricted regular FAQ rich results to well-known government and health sites years ago. The markup can describe content, but it is not an AI citation cheat code. 03 · fake brand mentions Do not manufacture forum posts, reviews or “best of” articles. Google specifically warns against chasing inauthentic mentions. Real reputation is harder to build because it is harder to fake — that is precisely why it is more valuable. 04 · 100 thin prompt pages Do not create a page for every imaginable AI prompt. Google warns that scaled pages made primarily to manipulate rankings or generative answers can violate its spam policies. One excellent page that covers the topic properly beats a graveyard of variations. 05 · “AI ranking guarantees” Nobody can guarantee that ChatGPT, Claude or Google AI will name you. OpenAI says ChatGPT Search ranking uses multiple factors and there is no way to guarantee top placement. Treat anyone promising guaranteed citations accordingly. Technical & content checks Seven practical AI-search checks to run before publishing more content These tests are deliberately practical. You can run them on a real business today and get a pass/fail result instead of another vague marketing score. 1 Google eligibility test Check the target URL in Google Search Console URL Inspection. It should be indexed, return a successful response, not be blocked by robots.txt, not carry noindex , and remain eligible to show a normal search snippet. Google states those Search requirements also underpin eligibility for its generative features. PASS: indexed + snippet eligible FAIL: blocked, noindex, error or canonicalised away 2 AI crawler access test Inspect /robots.txt and any firewall/CDN bot rules. If ChatGPT visibility matters, confirm OAI-SearchBot is not blocked. If Claude web visibility matters, check Claude-SearchBot and Claude-User . Do not assume allowing a training bot is the same thing as allowing search retrieval. User-agent: OAI-SearchBot Allow: / User-agent: Claude-SearchBot Allow: / User-agent: Claude-User Allow: / Example only. Keep private/admin paths protected and review the current crawler documentation before changing production robots rules. 3 Raw-content test View the page source or fetch the HTML without running JavaScript. Can you still see the service, location, prices where appropriate, proof, key answers and contact information? Google can render JavaScript, but JavaScript SEO is more complex, and other retrieval systems do not necessarily execute your site exactly like a browser. PASS: core facts exist in crawlable HTML FAIL: meaningful content appears only after client-side scripts run 4 Answer-quality test Pick the five questions a buyer actually asks before choosing you. For each question, find the paragraph that answers it. If the paragraph contains no number, example, decision rule, experience, source or useful distinction, it is probably commodity content. Example: “Does AI search reduce website clicks?” is weak if the answer says “sometimes”. It becomes useful when you explain that Pew observed 8% traditional-result clicks on pages with AI summaries versus 15% without, then explain the limitation: the study was a US panel in 2025 and does not prove every query behaves the same. 5 Entity consistency test Compare your website, Google Business Profile, key directories, review profiles and important trade/industry sources. The business name, service description, location, contact details and core claims should not contradict each other. An AI system can tolerate wording differences; it should not have to guess whether two conflicting descriptions refer to the same business. PASS: same real-world business story everywhere FAIL: conflicting category, location, phone or service information 6 Query fan-out coverage test Take one commercially valuable question and map the related questions somebody would need answered before making a decision. Do not create a separate thin page for each one. Build the strongest answers naturally into the right page or supporting content. 7 Measurement test Stop reporting “AI readiness” as a made-up score. Track what can actually be observed: Google’s generative AI Search Console reporting where available, Bing Webmaster Tools AI Performance, ChatGPT referral traffic, landing pages receiving AI referrals, brand mentions across repeatable prompt sets, and — most importantly — enquiries or revenue from those visitors. PASS: repeatable baseline + monthly comparison FAIL: screenshots of one favourable prompt used as proof Query fan-out in practice Why one strong page can surface for questions you never wrote verbatim Google says its generative systems can issue related searches to gather the information needed for an answer. The exact fan-out queries are generated dynamically and are not a public fixed list, but you can model the information need. Original buyer question: “Who is the best SEO company for a small Belfast business that needs more local leads?” Local proof: SEO company Belfast reviews and case studies Commercial fit: small business SEO costs UK Method: how local SEO improves Google Maps visibility Risk: how to choose an SEO company without a long contract Evidence: local SEO case study Northern Ireland Alternatives: SEO vs Google Ads for local lead generation Recommended publishing approach Do not turn those related questions into six doorway pages. Build one genuinely strong primary service page, link to a transparent pricing or process explanation where useful, publish real case-study evidence and create supporting guides only when they solve a distinct problem deeply enough to deserve their own URL. What “better content” actually means Original information beats another 3,000 words of paraphrased SEO advice Google’s current guidance repeatedly returns to unique, satisfying, non-commodity content. That is a useful standard because it forces a simple question: what can this page contribute that a generic AI answer cannot? Evidence a competitor cannot copy Real before/after Search Console data with dates and context Call or lead data with the client’s permission Original audits showing exactly what was changed Side-by-side tests of titles, layouts, pages or conversion paths Photos, screenshots and examples from real work Judgement a generic article avoids What you would not spend money on and why When SEO is the wrong channel Where industry studies disagree What changed since last year Which recommendations are proven, plausible or merely experimental Practical principle Evidence plus judgement is the moat. AI systems can summarise the same public documentation everyone else has read; they cannot legitimately manufacture a business’s real client outcomes, failed tests, local observations or decision-making process. Those are exactly the assets worth publishing. A useful benchmark A 100-point AI visibility audit for a real business This is not a Google score or a proprietary “ranking factor” claim. It is a practical prioritisation framework: if a business scores badly here, it has obvious work to do regardless of which AI engine is being discussed. Search eligibility & crawlability Indexing, 200 responses, robots, canonicals, snippet controls, rendered content. 20 pts Commercial content quality Clear services, buying questions, comparisons, limitations, pricing context where sensible. 20 pts Original evidence Case studies, data, screenshots, testing, first-hand experience, examples. 20 pts Entity & reputation consistency Business facts, reviews, major profiles, third-party mentions, real-world corroboration. 15 pts Topical coverage & internal linking Enough depth to satisfy adjacent questions without doorway-page duplication. 15 pts Measurement Search Console, Bing AI data, referral tracking, prompt sampling and conversion tracking. 10 pts The weighting is a practical prioritisation framework, not an external ranking formula. Its purpose is to expose obvious weaknesses before more money is spent on content or promotion. 90-day plan A practical 90-day plan for improving AI-search visibility Ninety days does not guarantee a citation or recommendation. The value of this sequence is that it fixes technical and content dependencies before spending heavily on promotion. Days 1–30 Fix access and the money pages Confirm Google indexing and snippet eligibility Audit OAI-SearchBot and Claude search/retrieval access Fix canonical, duplication and JavaScript rendering problems Rewrite the highest-value service pages around real buyer questions Add first-hand proof and remove generic filler Days 31–60 Build corroboration Make core business facts consistent across important profiles Strengthen review acquisition and response processes Publish real case studies and original examples Earn legitimate industry/local mentions Build internal links around true topic relationships Days 61–90 Measure and iterate Check Google generative reporting where available Review Bing Webmaster Tools AI Performance Segment AI referrals in analytics Repeat the same commercially relevant prompt set Improve pages using gaps found in real queries, not guesses Limits of the evidence There is still a lot we do not know — and pretending otherwise makes the advice worse Correlation is not causation. Large citation studies can show which pages appear together with AI citations, but they cannot reveal the complete internal ranking systems of Google, OpenAI or Anthropic. Results change quickly. Ahrefs reported 76% top-10 overlap in a 2025 AI Overview study, then its larger 2026 update found roughly 38% overlap. That is a huge change in less than a year and a good warning against treating last year’s correlations as permanent rules. AI visibility is query-sensitive. An informational prompt, a local recommendation, a product comparison and a breaking-news question can rely on very different retrieval sources. A citation is not automatically a lead. Pew’s click data is a reminder that AI answers can satisfy the user without sending traffic. Brand exposure, assisted conversions, direct searches and eventual enquiries matter alongside referral clicks. The practical conclusion Build for retrievability, usefulness, corroboration and commercial trust . Measure everything you can. Treat everything else as a hypothesis until your own data proves it is worth keeping. Frequently asked questions AI-search visibility questions answered clearly Can I guarantee my business appears in ChatGPT Search? No. OpenAI explicitly says there is no way to guarantee top placement. You can improve eligibility and relevance by keeping the site public and accessible to OAI-SearchBot, then improving the quality and authority of the information available about the business. Does SEO still matter for Google AI Overviews and AI Mode? Yes. Google says its generative Search features are rooted in its core Search ranking and quality systems. The 2026 Ahrefs data also shows overlap with organic rankings, although the majority of cited URLs in that study were not top 10 for the original query. Should I add llms.txt? Not for Google rankings. Google says it ignores llms.txt. There may be reasons to maintain one for other systems that choose to support it, but it should not be sold as a Google AI visibility tactic. Does FAQ schema help me appear in AI answers? There is no official evidence that FAQ schema is a special AI ranking factor. Use structured data when it accurately describes visible content and supports a legitimate Search feature. Do not add it everywhere simply because an “AEO checklist” told you to. What is the most valuable thing I can add to a page? Something original and useful: a test, a dataset, a real case study, a decision rule, photographs from the work, a comparison based on experience, a pricing breakdown, or a strong opinion you can defend with evidence. Commodity summaries are increasingly easy to generate and therefore increasingly hard to differentiate with. How do I measure AI traffic? Track referral traffic from AI services in analytics, use Google’s generative AI Search Console reporting if it is available to your property, review Bing Webmaster Tools AI Performance, and keep a repeatable prompt set for brand/citation sampling. Then tie those visits and mentions back to calls, forms and revenue where possible. Sources Primary documentation first, large studies second These are the sources behind the factual claims above. Where this guide uses a prioritisation framework or recommendation, it is presented as practical guidance rather than an official ranking factor. Google Search Central — Optimizing your website for generative AI features on Google Search (2026) Official guidance covering RAG, query fan-out, technical eligibility, llms.txt, chunking, schema, content quality and measurement. Read Google’s guide Google Search Central — Succeeding in AI Search (2025) Official guidance on unique content, page experience, crawlability and structured data. Read the Google Search Central post Google Search Central — Generative AI performance reports in Search Console (2026) Official announcement of dedicated generative-AI visibility reporting rolling out to a subset of sites. Read the announcement OpenAI — ChatGPT Search / Publishers and Developers FAQ Official guidance on OAI-SearchBot, discoverability, citations and the fact that top placement cannot be guaranteed. Read ChatGPT Search guidance Anthropic — Web crawler documentation Official explanation of ClaudeBot, Claude-User and Claude-SearchBot and how robots.txt controls each purpose. Read Anthropic’s crawler guide Ahrefs — 4M AI Overview citation study (2026) Large independent study reporting 37.1% organic top-10 overlap for cited URLs in its updated dataset. Read the Ahrefs study Pew Research Center — Google users and AI summaries (2025) Observed browsing behaviour from 900 US adults and 68,879 Google searches, including click rates with and without AI summaries. Read the Pew analysis Microsoft Bing — AI Performance in Bing Webmaster Tools (2026) Official announcement of citation and URL visibility reporting across Microsoft AI experiences. Read the Bing announcement Black Rhino Build a site that deserves to be retrieved, cited and chosen Black Rhino can audit search eligibility, AI crawler access, commercial content, entity consistency, traditional search visibility and measurement — then focus the work on changes that can be supported by evidence. Contact Black Rhino