AI visibility audit checklist

Can AI crawlers reach the site?

What to check: the robots.txt file, any firewall or CDN rules that block unusual crawlers, and whether the important words on each key page exist in the HTML.

How: open /robots.txt and look at the rules for OAI-SearchBot, GPTBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot and Google-Extended. Then load a key page with JavaScript disabled, or view the source, and confirm the service, location and who it is for are still readable.

What good looks like:

• OAI-SearchBot is allowed. OpenAI uses it to surface sites in ChatGPT search features. Blocking it means those answers will not show the site, though a navigational link can still appear.
• PerplexityBot is allowed if you want the site to be eligible for Perplexity search results.
• Claude-SearchBot is allowed if you want Anthropic to index the site for search. Anthropic documents this separately from ClaudeBot, which is used for model training.
• A CDN or firewall is not challenging those crawlers, and key text is in the HTML.

A note on Google-Extended and GPTBot: these control training or, in Google's case, some systems outside Search. Google says robots.txt rules for Googlebot are the control for AI features in Google Search. Blocking Google-Extended does not remove a page from AI Overviews or AI Mode. The site-level switch for those features is the Search generative AI control in Search Console.

Is it indexed in Google and Bing?

What to check: whether the important URLs are in each index, and whether a sitemap has been submitted.

How: in Google Search Console, use the Page indexing report and the URL Inspection tool. In Bing Webmaster Tools, check the sitemap status and submit an XML sitemap. Consider IndexNow so Bing is told when a URL changes. Also run a private-window site: search on both engines.

What good looks like: the service, location and guide pages are indexed; the sitemap lists the live URLs you care about and both tools have processed it; and IndexNow, if you use it, has a valid key file. IndexNow notifies participating search engines of a change; it does not mean a page will be crawled or indexed.

Google says a page must be indexed and eligible to be shown with a snippet before it can appear as a supporting link in AI Overviews or AI Mode. Google's longer guide to generative AI features adds one condition: the site must be included in Search generative AI features in Search Console. Inclusion is the default setting, so the check is that nobody has switched it to exclude.

Do the pages answer real questions?

What to check: whether each service and each distinct location has a page, and whether that page states the answer in the opening sentences.

How: list the services and places a buyer would ask about. For each one, open the page you would want cited and read only the first screen of text.

What good looks like: one page per service, and a separate page where a location has its own facts; the first sentences say what the service is, who it is for, and where it is offered; names and places are in ordinary text, not only in images or PDFs; and headings sound like the questions people ask.

A page that never states the answer is a weak source, even if it can be found. See AEO and GEO.

Is the structured data accurate?

What to check: the structured data on the home page, service pages and any FAQ.

How: view the page source, or use Google's Rich Results Test, and read the markup as if it were a public claim.

What good looks like: organisation markup (schema.org Organization, or the correct type for the business) with a sameAs list of official profiles; Service markup only where a real service is described; FAQPage markup only for questions and answers that appear as visible text; and no review or rating markup unless the reviews are real and the markup is allowed. Do not invent an aggregate score.

Google's guide to generative AI features (last updated 10 July 2026) says structured data is not required for those features, and that there is no special schema you need to add. It remains useful for rich results, and it must match the visible text.

Do other sites say the same thing?

What to check: the business name, a short description, and the address, phone number and website across the profiles and directories the business already uses.

How: compare the home page with the official profiles (for a local business, start with Google Business Profile) and any directories already claimed. Note any conflict, including older trading names and closed locations.

What good looks like: the same facts, written plainly, in each place the business controls. Mentions elsewhere should be genuine. Google's guide warns that seeking inauthentic mentions “isn’t as helpful as it might seem”, and that its quality and spam systems apply to generative AI features too.

What do the answer engines say today?

What to check: a fixed set of real buyer questions, asked in the engines those buyers use, with a written record.

How: write ten questions a buyer would actually ask, in their words; ask them in a logged-out window in ChatGPT, Perplexity, and Google AI Overviews or AI Mode where they appear; record the date, the engine, who was named, and which URLs were cited; and repeat the same set about once a month.

What good looks like: a dated record, not an impression, showing whether the business is named, misdescribed or omitted, and which other organisations are cited instead. One run is a snapshot, not a score. Answers can change between sessions.

There is a separate guide on why a business may not show up in ChatGPT, and on how these engines choose sources.

A sample of how findings are written

Example only. This is the shape of the note. It is not a client's report, and it is not a result.

The note contains the public pages that were read; the questions that were asked and which answer engines were checked; what the public answers said; and the fixes we would write, each with a page address. We do not give the site a score out of 100. When we have a completed piece of work a client is willing to name, we will publish a short case study using the same method. Until then, this sample is the honest version of our proof.

What we will not promise

We use AI tools for research, content optimisation and AI-search monitoring, and we study how ChatGPT, Perplexity and Google AI choose what to cite. We do not promise a ranking, a traffic figure, or a mention in any answer engine. The checklist is a way to see the public record clearly, not a way to buy a place in an answer.

Questions

What's the difference between an AEO audit and a GEO audit?

They are closely related, and we treat them as one piece of work. AEO looks at whether a page is a clear source for a direct answer. GEO looks at whether the public pages are a fair source for a generative answer, which may name or cite the business. The checks above cover both. There is a short guide on how SEO, AEO and GEO fit together.

Do I need a paid tool?

Not for a first pass. The crawler, indexing and page checks can be done with the site itself and with Google Search Console and Bing Webmaster Tools. The prompt set can be run by hand.

How often should I re-check?

Monthly is a sensible rhythm. Answers can change between sessions.

Do you offer this for law firms?

The free check and the audit are not offered to law firms. Work with a firm is an ongoing retainer, and the firm approves all content before it goes live. That is explained on the law firms page. The same checklist can be adapted for dental and implant groups and for B2B software businesses, using only public pages.

Write to us

Write to hello@clearlyseen.co.uk for a written check. Name the business and the sites. Please do not send personal details about customers or patients.