AI visibility tracking means counting how often your website shows up in answers from AI search systems: as an impression in Google's AI Overviews and AI Mode, as a cited source in Microsoft Copilot, as a mention or link in ChatGPT, Perplexity or Gemini. There is no single number, only three methods that count different things. Two reports come free from the platforms: the generative AI performance report in Google Search Console (impressions only) and "AI Performance" in Bing Webmaster Tools (citations in Copilot and Bing). ChatGPT, Perplexity and Claude offer no report for site owners, so a fixed prompt set of your own fills that gap. Paid AI visibility tools run fixed prompt lists every day; according to their pricing pages they start at 29 US dollars (Otterly.AI), 80 euros (SE Ranking, 64 euros on annual billing) or 95 US dollars (Peec AI) a month, as of November 2026. AI answers vary by session, country and model, so what you get is a trend, not what any single user sees.
For my own site, that is the job of Sebastian, my AI employee for SEO and GEO. What to change so your site can be cited at all is in Optimizing Your Website for AI Search. All related posts are under SEO and AI Search.
What AI visibility tracking actually counts
You cannot add these rows up: a Google impression is not a Copilot citation, a hit in a tool's prompt list not a real search. Nobody serious can promise you rankings or citations either, and I don't.
Google: the generative AI performance report
According to Google, every website has had a separate Search Console report for generative AI features since August 31, 2026. It counts how often a link to your site was shown in AI Overviews or AI Mode, by page, country, date and device, with the usual 1,000-row limit, but no queries or clicks. Per Google's documentation, those clicks sit in the regular Performance report under the "Web" search type, without an AI label. An empty report means too few AI impressions or a site excluded from the AI features.
My own Search Console: in the three months from August 2 to November 1, 2026, just over 18 percent of all kevinwelter.com impressions in Google web search came from generative AI features, about 299,000 out of 1.64 million. How many led to clicks, the report does not say. More on AI Mode is in Google AI Mode: What It Is and How It Works, setup in Setting Up Google Search Console. What Google says about appearing in AI Overviews is in How to Rank in AI Overviews: Google's Answer.
Bing: AI Performance for Copilot
"AI Performance" in Bing Webmaster Tools has been available as a public preview since February 10, 2026, with "Total Citations", "Average Cited Pages", the grounding queries (the search phrasings the AI used to pull in your content) and citations per page. It covers Microsoft Copilot, AI-generated summaries in Bing and select partner integrations. Microsoft names two limits: the data is a sample of citation activity, and it does not measure ranking. Since June 2026 a preview adds intents, topics, citation share and period comparison; since October 2026 a beta adds "Topics Portfolio", which Bing says smaller sites may not see at all for lack of data.
For my own site, Copilot cited kevinwelter.com 183 times in the three months through September 16, 2026. The most-cited page reached 27 citations, the page with the most Google clicks only 8. Google ranking and AI citations are two metrics, and one does not follow from the other. Bing setup is in Submit Your Site to Bing: Setup and IndexNow.
ChatGPT visibility: your own prompt set
ChatGPT, Perplexity, Claude and Gemini give site owners no report of their citations. Measure yourself, with a setup you can repeat eight weeks later:
- Ten to fifteen fixed questions a prospect would really ask, at most one of them with your brand name.
- A fixed rhythm, for example monthly, each question at least twice per system, because answers vary.
- A fixed log per answer: date, system, country, signed in or not, are you mentioned, are you linked, who appears instead.
- Freeze the setup. If you swap questions, you start a new baseline and stop comparing before and after.
I have my prompt set run by Claude agents with web search; that measures Claude's web search, not ChatGPT. A measuring agent with project instructions about my business recognised who had commissioned the run (the full story is in What Is Generative Engine Optimization?). So: start measuring agents in an empty folder without project instructions, and when you ask by hand, do it logged out and without history where you can.
Two extra sources give hints, not proof of citation: server logs show fetches by ChatGPT-User, Perplexity-User or Claude-User that a person triggered, and web analytics show visits from an AI answer as a referral from that service, if the browser passes the origin along (otherwise as direct traffic). Both show usage, not whether you were mentioned.
Why every AI visibility number is fuzzy
Put this in every report:
- Answers vary. The same question returns different sources by session, country, language and model version.
- Personalization. For AI Mode, Google says that with history turned on, earlier searches and activity are taken into account. Your own query is not your customer's.
- Samples. Bing explicitly calls its report a sample, and a tool only queries a selection of prompts.
- Unknown population. A tool's visibility score or share of voice holds for its own prompt list. Useful as a trend inside the same tool, not as an absolute statement.
AI visibility tools compared
Paid trackers give you volume (many prompts, systems, daily runs, competitors), but no clicks and no platform citation counts. Google adds a caution: no third-party tool has access to its internal ranking or AI systems; use them if they help your workflow. Three examples, from each pricing page, retrieved November 6, 2026:
SE Ranking shows euros when retrieved from Germany. Upfront: I have a partnership with SE Ranking, and I earn a commission through my links to SE Ranking*. What I tested there was the AI search database, not the daily tracker. For Germany in October 2026, it listed kevinwelter.com with 28 prompts in which the AI answer links to my pages, all in Google's AI Overviews, none in ChatGPT, Perplexity, Gemini or AI Mode; details in SE Ranking Review: API, MCP and Real Numbers.
* Affiliate link: if you subscribe or buy credits through this link, I earn a commission. Your price stays the same. I have a partnership with SE Ranking.
Before buying, ask: is coverage documented for your language and market, and can you export the answer text per prompt? For around fifteen questions re-measured monthly, the two free reports plus your own prompt set are enough in my experience; that is a heuristic. A tracker starts to pay off with hundreds of questions, several markets or competitors.
How Sebastian measures for me
Sebastian runs the classic series from Search Console and Bing search performance next to the generative one from the AI report, "AI Performance" and the prompt set. Every figure comes from a saved tool result in his project folder. He's read-only in other people's accounts, and nothing goes live without me. He measures, I decide. How he works with and without a paid tool is in AI Employee With and Without an SEO Tool. Why an llms.txt file does not change these numbers on current evidence is covered in llms.txt: Hype vs. Evidence. To check whether your markup is clean, see Schema Markup Checker: Which Tool Tests What. How the numbers get into a weekly report is in SEO Reporting With AI: A Weekly Report.
Frequently asked questions
Can I track AI visibility for free?
Mostly. Search Console shows impressions in AI Overviews and AI Mode, Bing Webmaster Tools shows citations in Copilot. For ChatGPT and Perplexity, a prompt set of ten to fifteen fixed questions is enough.
Does Search Console show clicks from AI Overviews?
Not separately. The generative AI performance report counts impressions only. According to Google, clicks from AI features sit in the regular Performance report under the "Web" search type, without an AI label.
How do I measure ChatGPT visibility?
With a fixed prompt set, at least two runs per question, and a log. ChatGPT has no site-owner report like Bing's.
How accurate are AI visibility tools?
As accurate as their prompt list: useful as a trend, not a picture of what real customers ask. In my test, one database listed my site in Germany only in AI Overviews.
What to do next
Set up your baseline today: export the generative AI performance report from Search Console for three months, export "AI Performance" from Bing Webmaster Tools, write down ten questions and send each twice to ChatGPT, Perplexity and Copilot. File everything with a date and re-measure with the same setup in eight weeks. For the links between your pages, see the guide to internal linking. How to write pages that answer a question in full is in the guide to SEO content writing with AI.
If you want to track many questions, several markets or competitors, try the tracker from SE Ranking*: according to the vendor, the trial runs 14 days without a credit card, with 20 AI prompts a day. And if you'd rather not set up the tracking for your site yourself, get in touch.
* Affiliate link: if you subscribe or buy credits through this link, I earn a commission. Your price stays the same. I have a partnership with SE Ranking.
