What this audit checks
Nine blocks. The first eight are machine checks of your site and listings; the ninth asks the AI engines directly.
Free AI visibility audit
We ask ChatGPT, Gemini and Google AI Mode the questions your customers ask, check what they say about you, and show the fixes that matter. Free, about a minute, no email for the first result.
Nine blocks. The first eight are machine checks of your site and listings; the ninth asks the AI engines directly.
| # | Block | What we check | Why it matters |
|---|---|---|---|
| 1 | AI crawler access | robots.txt rules and the real server response to GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and others, including firewall blocks | If an AI system can’t fetch your pages, it can’t quote them |
| 2 | Readable without JavaScript | Whether your phone, address, services and prices are in the page text, not only in scripts or images | Many AI crawlers read raw HTML |
| 3 | Machine-readable business card | LocalBusiness / Organization markup: valid, complete, name matching your legal name | Errors here put the wrong name or hours into AI answers |
| 4 | Name, address, phone | The same details on your site, Google profile and directories | Contradictions are the most common source of wrong AI answers we see |
| 5 | Google Business Profile | Category, rating, review count, hours, website link | ChatGPT’s business cards in the UK and EU show exactly these numbers |
| 6 | Directories AI cites | The listings AI engines cite for your industry and country, e.g. Yelp and Zocdoc (US), Checkatrade (UK), Doctolib (FR/DE), Herold (AT) | AI cites directories more often than business websites |
| 7 | Pages buyers need | A page per service, prices, process, coverage area | AI can’t quote what you never published |
| 8 | Languages and freshness | Language versions, outdated pages and dates | In Vienna, German and English questions named almost completely different dentists |
| 9 | The AI answers themselves | Two customer questions for your category and city: ChatGPT ×3, Gemini ×2, Google AI Mode ×1, plus Google’s map results as a benchmark; and a question about your business itself in ChatGPT, AI Mode and Perplexity | Shows who gets named instead of you, and what AI gets wrong about you |
AI answers are not fixed. When we asked ChatGPT the same local question five times, the lists overlapped only partly, and about half of all businesses it named appeared in just one of the five runs. ChatGPT and Gemini rarely agreed with each other at all. BrightLocal found the same at scale: asked twice, AI engines repeat only 20–33% of the businesses they named (Sep 2026).
That is why the full report asks ChatGPT three times, Gemini twice and Google AI Mode once per question, and reports the result as a fraction: “named in 2 of 3 answers”. We name the city inside the question, because the tools we use set the country, not the city.
A single score would hide the two things you need to know: what is wrong on your side, and how often AI names you. Our technical checks are counts you can verify (“7 of 8 AI assistants allowed in”). AI mentions are fractions with dates. Every number in the report links to the evidence behind it, the raw answer, the file, the page.
We can’t see ChatGPT with your personal location, history or memory, or answers for paid subscribers. We don’t probe Copilot, Claude or Grok. Answers from AI APIs differ from what people see in the apps, in our tests most businesses named by the API never appeared in the app, so we read the app answers. And one day’s check is a snapshot, not a trend; we offer a free re-check after 30 days.
Businesses with customers in a place: dentists, med spas, real-estate agents, law firms, movers and home services. Agencies can run it for clients under their own brand, ask us.
We store the website you entered, your email if you give it, the report and the AI answers we collected. Report pages are not indexed by search engines and are deleted after 90 days. Marketing emails only with the separate, unticked box. The report is prepared automatically with AI from machine checks, and every number is checked against the evidence before it is shown.
Enter your website above. We ask ChatGPT the question your customers ask, “a good {category} in {city}”, three times, and show which businesses it named and whether you were among them, as “named in 1 of 3 answers”. Asking once yourself can mislead: the list changes between runs.
Yes. The first result needs no email. The full report needs a confirmed email address, nothing else, no card, no call. Marketing emails only if you tick that box separately.
It depends on the country. In the UK and the EU, the ratings on ChatGPT’s business cards matched Google profiles exactly in our checks. In the US, about half of the cards showed Yelp’s rating instead, marked with Yelp’s logo. The audit shows which source your card uses.
No stable rank exists. ChatGPT names roughly four or five businesses per answer and the list changes between runs. We report how often you were named across runs, not a position.
ChatGPT (3 runs), Gemini (2), Google AI Mode (1) and Google’s map results as a benchmark; Perplexity for the question about your business. Not checked: Copilot, Claude, Grok, and ChatGPT with your personal location or memory.
We store the website you entered, your email if you give it, the report and the AI answers we collected. Report pages are not indexed and are deleted after 90 days. We don’t sell data. You can ask us to delete your report at any time.