How to Measure AI Visibility: Check Whether ChatGPT and Co. Cite You
How to track whether ChatGPT, Perplexity and Google's AI features cite your website: with a fixed question set, a test log, web analytics and clear metrics.
In short: AI visibility cannot be read off like a ranking, but it can be observed reliably. You need a fixed set of questions that you ask regularly in ChatGPT, Perplexity and Google, a simple log and an analysis of the visits that come from AI services. What matters in the end is not the number of mentions but whether they lead to enquiries.
This guide explains why single tests mislead, how to build a measurement system in four steps and which metrics actually tell you something. Last reviewed: October 2026.
Why a single test proves nothing
AI answers are not the same for everyone. According to OpenAI, ChatGPT rewrites questions into targeted search queries before searching the web, using the approximate location from the IP address and, if memory is switched on, saved information about the person. The same question can therefore return different sources depending on location, account and timing.
On top of that, answers vary even for identical input. If you ask once and find your website or not, you have an observation, not a measurement. It only becomes meaningful through repetition with fixed rules.
A measurement system in four steps
1. Define a fixed question set
Choose 10 to 20 questions your customers actually ask before they hire a provider like you. Mix three types:
- Problem questions: for example "Why is my website not generating enquiries?"
- Solution questions: for example "How much does a website redesign cost?"
- Provider questions: for example "Which agency helps with SEO for dental practices?"
Phrase the questions the way people type them into a chat window, and do not change the set afterwards, so results stay comparable.
2. Test regularly under the same conditions
Ask your question set once a month in ChatGPT, Perplexity and Google. Ideally use an account without saved memories or a new chat, and note the location you test from. Ask each question at least twice, because individual answers vary.
3. Log the results
Record the same details for every answer:
- Date, tool and question
- Is your brand mentioned in the text?
- Is your website linked as a source, and if so, which page?
- Which competitors or other sources are named?
A simple spreadsheet is enough. What matters is that you follow the same process every month.
4. Analyse visits from AI services
According to OpenAI, ChatGPT automatically adds the parameter utm_source=chatgpt.com to links in its search results. Your web analytics can therefore show how many visits come from ChatGPT and which pages they land on. Referrals from perplexity.ai and similar services show up in the visit's source.
Google's AI Overviews work differently: according to Google, clicks and impressions from them are included in the Search Console Performance report under "Web", but not reported separately.
Which metrics matter
- Mention rate: for how many questions in your set are you mentioned? For example, 4 out of 20 questions gives 20 per cent.
- Citation rate: for how many of those is your website also linked? Only a link can bring visits.
- Pages cited: which of your pages are used as sources? This shows which content works.
- Visits from AI services: number and landing pages of visits from chatgpt.com, perplexity.ai and similar sources.
- Enquiries from those visits: the most important number. It shows whether the work pays off commercially.
Compare the figures month by month and against your competitors in the same question set, not against a general average.
What about AI visibility tools?
Specialised tools count mentions automatically across many questions and countries. They save time but do not replace your own question set: check which questions a tool asks, in which country, how often and whether it distinguishes between a mention in the text and a linked source. A single "AI score" without these details is hard to interpret.
Common mistakes
- One test as proof: a single answer says little; repetition is essential.
- A changing question set: asking different questions every month makes comparison impossible.
- Your own account with memories: personal information changes the results.
- Counting mentions only: without looking at visits and enquiries, the value stays unclear.
To improve your results, read our GEO guide and our article on ChatGPT SEO.
No time to run your own tests?
In our free AI visibility check, we ask three typical questions from your industry in ChatGPT and Perplexity and send you three priorities within two business days.
Frequently asked questions about measuring AI visibility
Not as precisely as a ranking, because answers vary by location, account and timing. With a fixed question set, regular tests under the same conditions and an analysis of visits from AI services, you can still track progress reliably.
Once a month is enough for most businesses. Ask each question at least twice and keep the question set unchanged.
ChatGPT automatically adds utm_source=chatgpt.com to links in its search results. You can filter by it in your web analytics.
They are included in the Performance report under "Web" but, according to Google, not reported separately.
The number of enquiries that come from visits via AI services. Mention and citation rates show the direction; enquiries show the commercial value.
Template for your test log
Create a spreadsheet with these columns and fill it in every month for your complete question set:
- Date
- Tool (ChatGPT, Perplexity, Google)
- Question (from your fixed set)
- Brand mentioned (yes or no)
- Website linked (yes or no, with page)
- Other providers named
- Note (for example notable wording or outdated information about you)
Once a month, add the number of visits from AI services and the enquiries they produced.
Where do you stand in ChatGPT and co.?
Send us your website address. We run the sample questions for you and send you three prioritised next steps. Free and without obligation.