How to Measure AI Visibility: Method and Metrics
In short
To measure your AI visibility, ask ChatGPT, Perplexity, Gemini and Google the same panel of questions every month, and record your mention rate and share of voice against competitors. Add referral traffic in GA4, Search Console and Bing reports, and AI crawler visits in your server logs.
Key takeaways
- AI answers change with every run: track how often you appear, not a position.
- A fixed panel of 20 to 50 questions, each asked several times, gives you a comparable baseline month after month.
- ChatGPT adds utm_source=chatgpt.com to its links, so its traffic shows up in GA4.
- In Search Console, clicks from AI Overviews and AI Mode are counted under the 'Web' search type.
- Server logs show which pages AI crawlers actually read.
Why is AI visibility hard to measure?
Because there's no fixed ranking: an AI writes a fresh answer for every request. Ask the same question twice and you may see different businesses mentioned.
A study by SparkToro and Gumshoe, based on nearly 3,000 responses, found that ChatGPT and Google's AI returned the same list of brands less than once in a hundred runs. The authors conclude that how often a brand appears is more stable than its position. The study wasn't peer-reviewed, but it matches what we see in practice.
So the right method combines two kinds of measurement: what AI tools say, and what those answers do for your website.
How do you build a question panel?
A panel is a fixed list of questions you ask AI tools at regular intervals. It's your barometer: it has to stay the same so results remain comparable.
- Pick 20 to 50 questions your customers actually ask, without your brand name: "which accountant for a small business in Brussels?", "best scheduling software for a car repair shop."
- Mix the intents: category questions, comparisons, problems to solve, plus a few questions about your brand to check that the information is accurate.
- Cover your languages. In Belgium, the same question in French, Dutch and English can produce very different answers.
- Query several engines: ChatGPT, Perplexity, Gemini, Google AI Overviews and Copilot.
- Ask each question several times and record every brand mentioned, with or without a link to its site.
This panel flows straight from your content strategy: questions where you're missing point to pages you need to create or strengthen, as our 20-point GEO checklist shows.
Which metrics should you track?
Four metrics are enough to manage your AI visibility. Calculate them by engine, by language and by question group.
| Metric | What it measures | Calculation |
|---|---|---|
| Mention rate | Is your brand named? | Answers mentioning you ÷ total answers |
| Citation rate | Is your site given as a source? | Answers linking to your site ÷ total answers |
| Share of voice | Your weight against competitors | Your mentions ÷ mentions of all tracked brands |
| Accuracy | Is what AI says about you correct? | Manual check: services, city, contact details |
You can add average position in lists, but only as a rough signal, given how much answers vary. Accuracy is often overlooked: an AI that credits you with the wrong service or an old address costs you customers.
How do you spot AI traffic in GA4?
By filtering traffic sources by referring domain. OpenAI states that ChatGPT automatically adds utm_source=chatgpt.com to its links, which makes them visible as a source in Google Analytics 4.
For other assistants, look at referring domains such as perplexity.ai, gemini.google.com, copilot.microsoft.com or claude.ai. The simplest approach is a custom "AI assistants" channel group that brings them together.
Then track what those visitors do: landing pages, engagement, contact requests. This traffic is often small in volume, but it arrives with a very specific question already in mind. Keep in mind that some visits may arrive without an identifiable source, for example from certain apps.
What do Search Console and Bing Webmaster Tools show?
They show the effect of AI-generated answers on your pages, but not every detail. They complement direct testing; they don't replace it.
- Google Search Console: according to Google, AI Overviews and AI Mode data is counted in the Performance report under the "Web" type. Rising impressions without rising clicks can point to answers that satisfy searchers without a click. For the Belgian picture, see our article on Google AI Overviews in Belgium.
- Bing Webmaster Tools: since February 2026, the AI Performance report, in public preview, shows how often your pages are cited in Copilot and Bing's AI summaries, and which ones.
Bing's report is currently one of the few official sources that actually counts citations. It doesn't cover ChatGPT or Perplexity.
What do server logs reveal about AI crawlers?
Your server logs record every visit, including those from AI crawlers. They show which pages these crawlers read, how often, and which ones they skip.
Separate two groups. Indexing crawlers such as OAI-SearchBot or PerplexityBot build the search index. User-triggered agents such as ChatGPT-User, Perplexity-User or Claude-User visit a page because someone is asking a question right then: according to OpenAI, ChatGPT-User does not crawl the web automatically.
A page these agents visit often is a page AI tools consult to answer questions. A page that never gets visited deserves an access check: robots.txt, CDN, content loaded through JavaScript. To understand how ChatGPT then picks its sources, read how to appear in ChatGPT.
How often should you measure, and what are the limits?
A simple rhythm works for a small business: monthly panel, weekly referral traffic, monthly logs, quarterly review. Measuring more often mostly captures noise.
The limits are real. Answers vary by model, language, location, user history and date. Some tracking tools query model APIs, whose answers can differ from the interface your customers use. No measurement is exact: only the trend over several months matters.
Finally, keep the link to business results. More mentions with no visits or inquiries means your panel questions need rethinking. The foundations of this approach are covered in SEO vs GEO.
Take the next step
Want a first reading right away? Our "Do AI assistants recommend your business?" demo queries four engines, including the real ChatGPT: try it on our AI demos page or on our SEO & GEO page. We then set up your tracking: free 30-minute consultation and a quote within 24-48 hours.
Frequently asked questions
How do I know if ChatGPT recommends my business?
Ask ChatGPT the questions your customers would ask, without naming your brand, and note whether you appear. Repeat each question several times, because answers change from run to run. The percentage of answers that mention you, tracked monthly, is a more reliable signal than a single test.
Can I see ChatGPT traffic in Google Analytics?
Yes. According to OpenAI, ChatGPT automatically adds the utm_source=chatgpt.com parameter to the links it shows. In GA4, those visits appear with chatgpt.com as the source. Other assistants can be spotted by their referring domain, such as perplexity.ai. Some visits may still arrive without any source information.
Does Search Console show AI Overviews?
Yes, but not separately. Google says impressions and clicks from AI Overviews and AI Mode are included in the Performance report under the 'Web' search type. You see the overall effect on your pages, but you can't isolate exactly how much comes from AI-generated answers.
How often should I measure AI visibility?
A monthly run of your question panel is enough for most small businesses. Referral traffic can be checked weekly in GA4, and server logs monthly. A quarterly review connects those numbers to the work you've done and helps you set the next priorities.
Are AI visibility tracking tools reliable?
They're useful for trends, not as absolute truth. Answers depend on the model, language, location and timing. Some tools query model APIs, whose answers can differ from the interface your customers use. Always cross-check these tools against your own tests and your actual traffic.
