AI visibility audit: The 15 signals that decide if ChatGPT can find you

An AI visibility audit systematically reviews the 15 signals that determine whether ChatGPT, Claude, Gemini and Perplexity can find, understand and reference your business — from AI crawler access to brand entity. All 15 explained, the most common findings, and how to run it for free.

An AI visibility audit is a systematic review of the signals that determine whether AI engines such as ChatGPT, Claude, Gemini and Perplexity can find, understand and reference your business. Geoa's audit measures 15 technical dimensions across four groups — from AI crawler access to brand entity — and can be run for free in 60 seconds. Here are all 15 signals, the most common findings and how to interpret the results.

Why an audit — and not just a SEO check?

A classic SEO check measures whether Google can crawl and rank your site. An AI visibility audit asks a different question: can an answer engine use you as a source? The need for this is driven by the fact that the answers are eating into the clicks: Ahrefs’ analysis of 300,000 keywords showed a 34.5 per cent lower click-through rate for the top position when an AI answer appears at the top — and 58 per cent in the follow-up analysis in December 2025. The battle is shifting into the answer itself, where different signals apply.

The good news is that these signals can be influenced: research by Princeton and Georgia Tech (KDD ’24) found a visibility boost of up to 40 per cent in generative search engines through targeted optimisation of precisely the kind of signals an audit uncovers. In other words: the audit doesn’t just find faults — it finds the levers that can demonstrably be turned.

An AI audit therefore looks for things that a SEO check would never measure — whether the AI crawlers are even allowed access, whether your content can be read without JavaScript, whether an llms.txt exists, and whether your company is a unique entity that the AI dares to cite. The difference is measurable: when we calibrated the measurement across 176 corporate domains (our own survey, August 2026), 36 per cent had a difference of at least 15 points between their traditional Google readiness and their AI readiness.

The 15 signals — group by group

The audit organises the 15 dimensions into four groups, each of which answers one question:

1. AI discoverability: can the engines get in?

• AI crawler access — are GPTBot, ClaudeBot and PerplexityBot blocked on robots.txt? This is the most heavily weighted signal, because without access, the rest is irrelevant.
• Server-side rendering — does your content require JavaScript? AI readers cannot run it.
• llms.txt — the curated overview of your business. A controversial signal, which we weight lightly but measure honestly.
• Security — HTTPS and sensible headers; basic trust.

2. Content extractability: can the answer be extracted?

• Quotable content — direct answers early on, question-based headings, short paragraphs.
• Basic page signals — titles, meta tags, H1 structure, canonical tags and ‘and’ tags.
• Internal links — can the crawler find its way to the rest of your content?
• Performance — slow pages are more likely to be abandoned by crawlers.

3. Entity citability: does the AI dare to mention you?

• Brand entity — is your business uniquely identified across profiles and directories, and does your site link to them?
• E-E-A-T quality — visible author, expertise and sources within the content.
• Schema recommendations — the schema types missing for your specific site type.

4. Structured data: are your facts machine-readable?

• Structured data — Organisation, FAQPage and other schema blocks.
• Platform optimisation — FAQ and Q&A formats that search engines pull up directly.
• Mobile optimisation — viewport and readability on small screens.
• Empirical content data — figures, statistics and concrete facts within the text; the elements that make a section worth quoting.

What does the audit typically find?

The patterns recur across industries. In our scan of 465 search-visible Danish corporate websites (our own survey, July 2026), 89 per cent were missing an llms.txt — and our analysis of robots.txt files showed that a significant proportion are blocking AI crawlers without realising it, typically because a security package or an outdated template has done so for them. Added to this are the two classic issues: content that is only accessible via JavaScript, and companies without a single machine-readable fact block.

It’s also worth knowing what an audit cannot do: it measures your foundations, not your reputation. In our analysis of 576 AI responses (our own survey, July 2026), the technical aspects were a strong predictor of whether a site was cited as a source — but told us virtually nothing about whether the company was recommended. Recommendations require reviews, mentions and third-party validation on top of the fundamentals. A good audit will tell you both of these things honestly.

Finally, there’s a pitfall you need to be aware of before comparing audits across different tools: some ‘audits’ are actually sales pitches that flag everything as red to sell a subscription. Check whether the tool shows you what it has measured (the specific robots.txt line, the missing schema), whether it dares to show green fields — and whether it is honest when something could not be measured. A measurement you cannot verify is not a measurement.

How do you run the audit for free?

You don’t need an account, a credit card or any technical knowledge:

1. Enter your domain into the free test.
2. Wait around 60 seconds — the scan measures all 15 dimensions of your live site.
3. Read the results from the top: the two indicators (Google visibility and AI visibility), your overall score with a grade, and then the dimensions one by one, with the most important fixes first.

Each fix comes with a specific instruction — and if you use WordPress or Shopify, the platform can implement the technical fixes directly on the site and measure the impact week by week.

How do you interpret the results?

Start with the two indicators at the top. Your Google visibility summarises the classic technical search signals; your AI visibility summarises the AI-specific ones. If there’s a big gap between them, you’ll know straight away which channel is lagging behind — and it’s typically the AI indicator that reveals the low-hanging fruit, because very few people have ever looked at its signals.

Next comes the overall score, graded from A to F, and all 15 dimensions shown as bar charts. Don’t chase a score of 100 across the board: look for the dimensions where the bar is short and the weighting is high — these are the ones the prioritisation places at the top of the action list. Two or three targeted fixes typically make more of a difference than ten minor ones, and because the audit can be run again for free, you can see the effect of each fix for yourself, rather than just taking it on faith.

Frequently asked questions

How long does an AI visibility audit take?

The free version takes around 60 seconds — it scans your live site across all 15 dimensions. A manual review of the same signals would typically take half a day to a full day; that’s exactly what automation saves you.

What’s the difference between an AI visibility audit and a SEO audit?

A SEO audit measures against Google’s blue links; an AI visibility audit measures against the search engines. The fundamentals overlap (a technically sound site), but the AI audit measures signals that SEO audits do not recognise: AI crawler access, llms.txt, citability and brand entity. See the difference between the two channels in the guide ‘Visible on Google’ — and in ChatGPT.

How often should I run an audit?

After every major site change, and otherwise monthly. The signals change when you switch themes, plugins or CMS — and you’ll only detect a blocked AI crawler by measuring.

Can I fix the issues myself?

Most of them, yes. robots.txt and llms.txt fixes can be made in minutes using the free tools; schema and citability require a little more effort. The audit prioritises fixes by impact, so you start in the right place — whether you do it yourself or let the platform do it.