The AI visibility audit: a 40-point checklist for UK businesses

When Google shows an AI Overview, the pages it cites earn roughly 2.1% click-through, while pages on the same results page that aren’t cited get about 0.9%. That’s Seer Interactive’s analysis of 5.47 million queries, and it means being named in the answer is now worth more than ranking just below it. Meanwhile 54% of UK adults now use AI tools such as ChatGPT, Copilot or Gemini, according to Ofcom.

So the question for most established businesses isn’t whether AI search matters. It’s whether the assistants can find you, understand you, quote you and trust you. This AI visibility audit answers that in 40 checks across five areas. Each one tells you how to test it and what good looks like. Run it in an afternoon, score yourself, and you’ll know exactly what to fix first.

On this page

• How to run the audit
• Access: can AI systems reach your site? (checks 1 to 8)
• Entity: does AI know who you are? (checks 9 to 16)
• Content: can AI quote you? (checks 17 to 26)
• Off-site: does anyone else vouch for you? (checks 27 to 34)
• Measurement: can you see what’s happening? (checks 35 to 40)
• Scoring and what to fix first
• FAQs

How to run the audit

You need a browser, Google Search Console, Bing Webmaster Tools, GA4 and a spreadsheet. Score every check 0 (missing), 1 (partly there) or 2 (done properly). The maximum is 80.

Be honest with the scoring. A check you “sort of” pass is a 1, not a 2. The point is to find the gaps, not to feel good about the ones you’ve already closed.

Access: can AI systems reach your site?

Access comes first because nothing else matters if the assistants can’t read your pages. These are also the quickest fixes on the list, often a single line in a file.

1. Your robots.txt lets AI search bots in. Open yoursite.co.uk/robots.txt and check that OAI-SearchBot (ChatGPT search), Claude-SearchBot and PerplexityBot aren’t disallowed. Perplexity says plainly that PerplexityBot is what surfaces and links sites in its results. Block it and you’ve opted out of its answers.
2. You’ve made a deliberate call on training bots. GPTBot, ClaudeBot and the Google-Extended token govern whether your content can be used to train models. Blocking them is a legitimate choice, but it’s a separate decision from search visibility, and blocking Google-Extended doesn’t remove you from AI Overviews, which run on Googlebot. Write the decision down so nobody “fixes” it by accident.
3. Your firewall or CDN isn’t quietly blocking them. Cloudflare and other providers offer one-click AI bot controls, and they’re easy to switch on and forget. Cloudflare’s bot reference lists the user agents to check. If robots.txt says yes and the firewall says no, the firewall wins.
4. Key content is in the HTML. Prices, service descriptions and FAQs locked inside images, PDFs or content that only appears after JavaScript runs are hard for many crawlers to read. View the page source: if the text you care about isn’t there, assume a bot can’t see it either.
5. Your important pages are indexed in Google and Bing. AI Overviews draw on Google’s index and Microsoft Copilot draws on Bing’s. Check the coverage reports in Search Console and Bing Webmaster Tools for your money pages.
6. You haven’t restricted snippets by mistake. A nosnippet tag or a tight max-snippet setting limits how much Google can show, AI features included. Search your templates for both.
7. Your XML sitemap is clean and dated. Every indexable page is listed, nothing redirected or broken is, and lastmod dates are real rather than all set to today.
8. Bots are actually visiting. Ask your developer or host for server logs and filter for the AI user agents above. No visits in a month usually points to a block you haven’t found yet.

Entity: does AI know who you are?

Assistants recommend businesses they can identify confidently. If your name, location and services are described five different ways across the web, the model has to guess, and it may guess wrong.

9. Your name, address and phone number match everywhere. Compare your website footer, Google Business Profile, Companies House record, LinkedIn page and the main directories in your sector. Small differences (“Ltd” here, “Limited” there, an old address on one listing) add doubt.
10. Organisation schema is on your site, with sameAs links. Organization or LocalBusiness markup should state your legal name, logo, address and links to your official profiles. Test it in Google’s Rich Results Test.
11. Your Google Business Profile is complete. Categories, services, opening hours, photos and regular review responses. For local “near me” style questions, it’s one of the first things an assistant leans on.
12. Your About page states facts, not slogans. When you were founded, where you’re based, who runs the business, what you do and for whom. “Passionate about excellence” gives a model nothing to work with. “Founded in 2009 in Leeds, 40 staff, specialists in commercial property law” does.
13. You know where you stand on Wikipedia and Wikidata. Wikipedia is ChatGPT’s second most-cited domain in Ahrefs Brand Radar data compiled by Obsurfable. Most businesses won’t meet Wikipedia’s notability rules, and paying someone to write your page usually ends in deletion. A factual Wikidata entry is often the more realistic step.
14. The assistants describe you accurately. Ask ChatGPT, Gemini, Perplexity, Copilot and Claude “Who is [your business]?” and “What does [your business] do?” Note anything wrong, out of date or missing. Errors usually trace back to an old listing or page that still says the wrong thing.
15. Your services have consistent names. If your site calls it “commercial cleaning”, your listings say “office cleaning” and your LinkedIn says “facilities services”, decide which you want to be known for and use it everywhere.
16. Your experts are visible. Named people with real job titles, short bios and LinkedIn profiles, attached to the content they’re responsible for. Anonymous content is harder for anyone, human or machine, to trust.

Content: can AI quote you?

This is where most established sites lose out. They have plenty of pages, but few of them contain a clean, specific answer that an assistant can lift and attribute.

17. Each section opens with a direct answer. The first 40 to 60 words under a heading should answer the heading’s question on their own. If an assistant took just that paragraph, would it make sense?
18. You answer the questions buyers actually ask. List 20 to 50 questions your customers put to you before buying, phrased the way they’d type them into ChatGPT (“best accountant for a limited company in Warrington”, “how much does a commercial roof survey cost”). Check you have a page that answers each one.
19. You have comparison, pricing and “how to choose” pages. Seer found AI Overviews on around 95% of comparison queries but only about 5% of transactional ones. If you don’t publish the comparison, someone else’s version gets quoted.
20. Facts are specific and dated. “Most projects take 6 to 8 weeks (2026 figures)” beats “projects are completed quickly”. Models favour statements they can repeat without hedging.
21. You publish something nobody else can. Your own data, case studies with real numbers, worked examples in pounds, or a clear opinion backed by experience. Rewritten versions of page one of Google are exactly what AI already has plenty of.
22. Genuine FAQs are marked up. Where a page has a real FAQ section, add FAQPage schema that matches the visible questions word for word.
23. Pages show when they were last updated. A visible “last updated” date, backed by an actual update, signals the information is current.
24. Specs, prices and comparisons sit in tables. Tables are easy for people to scan and easy for machines to parse.
25. Your headings follow a clean hierarchy. One H1, logical H2s and H3s, and headings that say what the section answers rather than “Overview” or “More information”.
26. Location and service pages carry real local detail. Areas covered, local projects, local reviews. A Manchester page that’s your Leeds page with the city swapped won’t convince anyone.

Off-site: does anyone else vouch for you?

Assistants don’t take your word for it. They look for corroboration, and the sources they cite most are often not your website at all.

27. You have recent, plentiful reviews on the platforms that matter in your sector. Google for local services, Trustpilot for consumer brands, G2 or Capterra for software, sector-specific sites where they exist. Respond to them, good and bad.
28. You’re present where AI looks. Reddit, YouTube, LinkedIn and Wikipedia top Peec AI’s analysis of 30 million citations across ChatGPT, Google AI Mode, Gemini, Perplexity and AI Overviews. You don’t need to be everywhere, but you should know which of these your buyers use.
29. Trade press and associations mention you. Membership listings, awards, interviews and commentary in your industry’s publications all count as independent corroboration.
30. Sector directories are accurate. The niche directories in your industry often outrank you for your own category, and assistants read them.
31. You appear on credible “best of” and comparison lists. Earned, not paid for. If you’re absent from every list in your category, find out why.
32. Your YouTube videos have proper titles, descriptions and transcripts. YouTube holds an outsized share of Perplexity and AI Overview citations in the same Brand Radar data, and transcripts are what gets read.
33. Your LinkedIn company page is complete and active. Accurate description, specialisms, location and regular posts from the company and its people.
34. You monitor unlinked mentions. Mentions of your brand without a link still shape how models describe you. Track them monthly, and correct anything inaccurate at the source.

Measurement: can you see what’s happening?

You can’t improve what you don’t measure, and AI visibility is easy to measure badly. Referral traffic shows only the people who clicked, not the far larger number who read an answer that named you, or didn’t.

35. GA4 separates AI referrals. Create a custom channel group or exploration that matches session source against a pattern such as chatgpt|openai|perplexity|claude|gemini|copilot. Then you can see volume and conversion rate for AI visitors on their own.
36. You track a fixed set of prompts every month. The 20 to 50 buyer questions from check 18, run in each assistant on the same day each month, with the answers saved. Consistency matters more than volume.
37. You record citations, mentions and recommendations separately. Being linked as a source, being named in the text, and being actively recommended are three different outcomes. Treat them differently.
38. You know your share of voice against competitors. For each tracked prompt, who gets named? A rival that appears in 30 of 40 answers while you appear in 4 is your real competitor in AI search, whatever the rankings say.
39. You watch Search Console for the AI Overview pattern. Queries where impressions hold or rise but clicks fall are usually ones where an AI answer now sits above you. Advanced Web Ranking found 74.3% of searches with an AI Overview ended without an organic click, against 53.6% without one.
40. You ask leads where they found you. Add “ChatGPT or another AI assistant” as an option on enquiry forms and in sales calls. It catches the influence analytics will never see.

Scoring and what to fix first

Add up your score out of 80.

Whatever your score, fix things in this order:

1. Access (checks 1 to 8). Usually an afternoon of work, and nothing else counts until it’s right.
2. Entity (checks 9 to 16). A few days of tidying listings, schema and your About page.
3. Content (checks 17 to 26). The ongoing work: answer-first rewrites of your most important pages, then new pages for the buyer questions you don’t yet answer.
4. Off-site (checks 27 to 34). The slowest to build and the hardest for competitors to copy.
5. Measurement (checks 35 to 40). Set it up alongside step 1, so you have a baseline before you change anything.

FAQs

What is an AI visibility audit?

An AI visibility audit checks whether AI assistants such as ChatGPT, Gemini, Perplexity, Copilot and Google’s AI Overviews can access your website, identify your business correctly, quote your content and find independent evidence that you’re credible. It also checks whether you’re measuring how often they name you.

How often should I run an AI visibility audit?

Run the full audit once a quarter and track your prompt set monthly. AI platforms change how they search and cite sources regularly, so a site that scored well in the spring can lose ground by the autumn without changing anything.

Is an AI visibility audit different from an SEO audit?

It overlaps heavily, because AI Overviews and most assistants rely on pages that are already well indexed. The difference is emphasis: an AI visibility audit puts more weight on entity consistency, answer-first content, off-site corroboration and measuring mentions inside AI answers rather than just rankings.

Should I block AI crawlers to protect my content?

You can block training crawlers such as GPTBot and ClaudeBot without leaving AI search, because the search bots are separate. Blocking search bots such as OAI-SearchBot or PerplexityBot removes you from those assistants’ answers, which for most businesses costs more than it protects.

Need help acting on the results?

If your score has shown more gaps than you have time to close, that’s normal. Most of the businesses we audit pass the access checks and fall down on entity and off-site proof. You can read more about how we approach this in our complete guide to AI SEO and GEO, how LLM citations work, and what ChatGPT SEO involves.

Require AI SEO, GEO, or digital marketing and AI automation within your business? Book a free appointment with a senior consultant. Or see our AI SEO services.