JW2D OÜ

AI visibility tracking tools: what they can and cannot do

Yakiv Romanchuk, founder of JW2D · Published

Short answer

AI visibility tools come in three kinds: free official reports (AI Performance in Bing Webmaster Tools, your site analytics), paid trackers that ask AI a list of questions on a schedule and count mentions, and manual checks. None of them shows a “ChatGPT ranking” the way Search Console shows a Google position: AI answers change from run to run, and AI companies do not disclose real user prompts. So tracker data is an estimate that is useful for trends across many checks, not an exact measurement.

Three kinds of tools

Kind What it shows Limitations
AI Performance in Bing Webmaster Tools Citations of your site in Copilot, AI-generated summaries in Bing and select partners, cited pages, a sample of the queries AI used to find your pages Microsoft’s ecosystem only; ChatGPT, Gemini, Perplexity and Claude are not covered
Site analytics Visits from AI: ChatGPT adds utm_source=chatgpt.com to its links Only clicks are visible, not mentions without a visit. Google includes AI Overviews traffic in overall “Web” traffic in Search Console
Paid trackers Share of mentions for your brand and competitors across a question list, sources, sentiment, trends An estimate from a sample of questions; accuracy depends on the number of questions, repeats, language and how the AI is queried
Manual checks Real answers as a user sees them, with screenshots Time-consuming; few repeats mean high variance

Is AI visibility tracking accurate?

  • Answers are unstable. In research by SparkToro and Gumshoe (January 2026), the same brand recommendation question was asked of ChatGPT, Claude and Google’s AI in Search 60–100 times each. ChatGPT and Google’s AI returned the same list less than once in 100 runs, and the same list in the same order less than once in 1,000.
  • Answers are personalized. Ahrefs writes that answers depend on location, context and memory of past conversations, and that the same prompt run three times will likely produce three different answers.
  • The takeaway. A “ChatGPT position” from one check means very little. The meaningful metric is how often you are named across many questions and repeats, and how that share changes over time.

What is prompt volume?

Some trackers show “prompt volume”, an estimate of how often people ask AI a given question. Ahrefs says outright that AI companies keep prompt logs private and itself uses around 100 million synthetic prompts built from keywords and questions with known search demand. We have not found OpenAI, Google or Microsoft publishing how often their assistants are asked specific questions. So any “prompt volume” is an estimate or a conversion of search data, not real ChatGPT statistics.

About “historical data”

History in a tracker usually starts when the questions began to be tracked; ChatGPT does not publish past answers. If a tool promises data going back years, ask from when and on which set of questions it was collected.

How to choose a tool

  1. Platforms. Which AI does it track: ChatGPT, Gemini, Perplexity, Claude, Copilot, Google AI Overviews and AI Mode?
  2. Languages and countries. Does it support questions in Estonian and Russian and a location in Estonia? For a local business this is the deciding point.
  3. Your own questions. Can you upload your customers’ real questions, not just generated ones?
  4. Repeats. How many times and how often is each question asked? More repeats mean less random variance.
  5. Query method. Through an API or through the interface users see? Ask the vendor how results might differ.
  6. Sources. Does it show which sites the answers cite? That is the main thing you will work on afterwards.
  7. Export. Can you download raw answers to check them yourself?

What a small business should use

A free setup is enough to start: a fixed list of 10–20 questions, a monthly manual check across five AI assistants with a few repeats, the AI Performance report in Bing Webmaster Tools and a utm_source=chatgpt.com filter in analytics. A paid tracker makes sense when you have dozens of questions, several languages and need regular competitor comparisons. At JW2D we send clients a monthly report on ChatGPT, Gemini, Perplexity, Claude and Copilot answers; see AI search optimization. What to check is described in AI visibility.

Sources

  1. Microsoft Bing. Introducing AI Performance in Bing Webmaster Tools Public Preview
  2. Google Search Central. AI features and your website
  3. OpenAI Help Center. Publishers and Developers — FAQ
  4. SparkToro. New research: AIs are highly inconsistent when recommending brands or products (January 2026)
  5. Search Engine Journal. AI Recommendations Change With Nearly Every Query: SparkToro (30 January 2026)
  6. Ahrefs. You Can’t Track AI Like Traditional Search (12 September 2025)

FAQ

What is the best AI visibility tracking tool?

There is no universal best. Choose by platforms, languages and countries, the ability to upload your own questions, number of repeats and whether the tool shows the sources behind answers.

Is AI visibility tracking accurate?

It is an estimate. AI answers change from run to run and are personalized, so only the trend in share of mentions across many questions and repeats is reliable, not a position in one answer.

What is prompt volume?

An estimate of how often people ask AI a given question. AI companies do not disclose prompt logs, so these figures are built from synthetic prompts or search data.

Are there free AI visibility tools?

Yes: the AI Performance report in Bing Webmaster Tools, Google Search Console, a utm_source=chatgpt.com filter in analytics and manual checks of AI answers.

Does Google Search Console show AI Overviews traffic?

It is included in overall “Web” traffic in the Performance report; Google does not report it separately.

Want AI to recommend your business?

We will check what ChatGPT, Gemini and Perplexity say about your niche today and show you where to start.

Discuss your project About the service WhatsApp

More articles