← Journal · AI visibility · 3 min

What is an AI visibility audit?

5DAYS studio1 NOV 2026
An adjustable desk inspection lamp shining on a blank technical drawing
An AI visibility audit checks whether AI tools can read your site, understand what you do and find others confirming it, and how you appear in their answers today. There's no standard definition, so ask what's included. A useful audit ends in a prioritised list of fixes and a baseline, not a single score.

What should it check?

A good audit covers four layers, in this order, because a problem in one makes the next less meaningful.

LayerQuestionExamples of checks
AccessCan AI systems reach your pages?robots.txt rules for OAI-SearchBot and PerplexityBot, firewall and CDN settings, noindex and snippet rules, text available without scripts
UnderstandingIs it clear what you do?Plain descriptions of service, audience and area, facts in text instead of images, accurate structured data
ConfirmationDo others say the same?Profiles, directories, reviews, press, consistent name and details
PresenceWhat do answers say now?Fixed questions in several tools, sources cited, how you're described, errors

What should it deliver?

  • A baseline. The questions, the tools, the date and the raw answers, so that a later measurement can be compared.
  • A list of findings with evidence. Each with a screenshot, URL or setting, not a general remark.
  • A priority order. What to fix first, who does it and what effort it takes.
  • An honest split. Documented requirements on one side, assumptions on the other.
  • A note on what can't be known. Answers vary and the systems are closed.

What can't an audit prove?

It can't tell you why a specific answer did or didn't name you. OpenAI says only that it ranks using multiple factors, and Google states that no third-party tool has access to its internal ranking or AI systems. So an auditor sees the input and the output, not the inside. And since answers keep changing, one round of testing is a snapshot. Treat conclusions drawn from it as hypotheses.

Which first-party data should it use?

Start with what the platforms report themselves. It's limited, but it isn't guesswork.

  • Google. The Generative AI performance report in Search Console shows impressions in AI Overviews and AI Mode.
  • Microsoft. AI Performance in Bing Webmaster Tools reports citations in Copilot, Bing's AI summaries and selected partners.
  • OpenAI. ChatGPT adds utm_source=chatgpt.com to referral links, so your analytics can show visits.
  • Your own leads. Ask new customers how they found you. It's simple and often overlooked.

How do you spot a weak audit?

  • One score out of 100, with no explanation of how it's calculated
  • A promise that the score predicts mentions or revenue
  • Questions that include your brand name only, which says little about whether you're found
  • No raw answers, so nobody can check the work
  • Recommendations that all lead to one product or to a pile of new articles
  • Claims about internal AI ranking data

Should you do it yourself or hire someone?

Do it yourself first if you have a small site and someone with an afternoon free. Use our 15 checks, test questions and the free scan. Get help when you can't see or change crawler settings, the site is large, or you want the measurement repeated without tying up your team.

At 5DAYS, a Complete or Signature website includes an AI Visibility Scan as baseline and delivery check, and our Visibility & Growth plan (€299 a month, excluding VAT) includes periodic scans, a quarterly report and one defined optimisation a month. We don't guarantee results. They depend on external platforms.

02 / Questions

Good to know.

Want to know how AI reads your site?