Measure AI citations without overstating the data

Start with important prompts

A prompt is a question your buyer would ask an assistant, not a keyword. “Which project management software is best for a remote team?” is an invitation; “Project management software” is not it. Write your own or assemble them from your Search Console queries with one click, which is proven demand and not guesswork. Each call can include a market and a language, because the answer in Germany is not the answer in Texas.

What each check records

Each run asks ChatGPT, Perplexity, Gemini, and Google AI Overviews for your prompt and stores the full answer, each citation with its position, whether your brand was mentioned, and which competing domains appeared instead. Nothing is paraphrased: you can expand on the exact response text a buyer would have read.

Read the results as frequencies, not judgments

The same engine can answer the same question differently an hour apart. For this reason, results are reported as frequency, reported in 4 of the last 5 trials, with honest labels for early, directional, and stable data. A single “uncited” is noise; A stable 0 out of 5 on a request for money is a result. Weekly or daily scheduled runs create this history without you having to think about it, and the visibility trend chart shows the direction.

The crawl connection: Why You Win or Lose

Because the tracker sits alongside your crawl, it can explain results rather than just report them. The three diagnostics you should check first: a cited page that your robots.txt is blocking for AI crawlers, a reply page whose content only exists after JavaScript is executed, and a high equity page that is never cited. For lost answers, Where Competitors Win shows the prompts, the domains they beat, and the angle they use, and in a gap description, the content plan is built to take the answer back.

Make your website easy to cite

The AI ​​Readiness Score evaluates eight signals, from site scanner access and structured data to server rendering and entity identity. The handy checklist: Allow the desired AI crawlers (GPTBot, ClaudeBot, PerplexityBot and friends, with ready-to-copy robots.txt blocks in both directions), publish a llms.txt ordered by your internal link ownership so assistants can find your best pages, generate and validate one agents.json and make sure that important content is rendered on the server side. Bot coverage then interprets your server logs to show which AI crawlers are actually fetching which pages, and AI referral analysis shows the visit assistants sent to you in GA4.