How to check whether AI assistants mention your business
You can measure some of this, sample more of it, and none of it completely. This guide separates the data the platforms actually provide from the spot checks you can run yourself, and says what each one can and cannot tell you.
Last updated . Written by Zack Wilde. Prices quoted are Wilde Digital's own published starting rates as of September 12, 2026; they are not market averages.
The short answer
Use three sources, in this order: the platform reports that now exist (Google Search Console and Bing Webmaster Tools both report AI-feature visibility), your own analytics for visits referred by assistants, and a small, repeatable set of manual prompts. Treat the reports as the most reliable, the referrals as partial, and the prompts as a sample — useful for spotting problems, not for claiming market share.
The platform reports
- Google Search Console. Google’s generative AI performance report shows impressions from AI Overviews and AI Mode. It reports impressions, not AI-only clicks, leads or revenue.
- Bing Webmaster Tools. Bing’s AI Performance report (public preview since February 2026) shows how often your pages are cited in generative answers across Microsoft’s AI experiences, which URLs, and sampled grounding queries. Bing notes that the figures do not indicate “ranking, authority, or the role of any page within an individual answer.”
- Bing Topics Portfolio. A beta added in October 2026 groups those queries into themes and lists opportunity queries where other sites are cited and yours is not. It only appears if there is sufficient data for the site, and Bing asks that its numbers be treated as directional.
Both reports need a verified site. Set them up before you need the data; neither reconstructs history from before you started looking.
Referrals in your own analytics
When an assistant shows a link and someone clicks it, the visit can arrive with a referrer or tracking parameter that names the assistant’s domain, such as chatgpt.com, perplexity.ai or copilot.microsoft.com — but not always, so treat what you can identify as a minimum. Group those visits, then follow them to inquiries.
- Report AI referrals by landing page and by inquiry, not just as a visit count.
- Expect undercounting: some apps and privacy settings strip the referrer, and a person who reads an answer and later types your address appears as direct traffic.
- Never reassign the unknown group to AI because it would make a better story.
Manual prompt checks, done properly
- Write ten to twenty questions a real buyer would ask, phrased the way they would phrase them, without your business name.
- Run them on each assistant you care about, logged out or in a fresh session where possible, and record the date.
- Record what you see: whether you are mentioned, whether you are linked, which page is cited, and what the answer says about you.
- Repeat monthly with the same questions. Answers vary between runs and users, so one result means little; a pattern over months means more.
- Fix what is wrong, not what is absent. An answer that states your prices or services incorrectly points to a specific page to correct. Absence alone has many possible causes.
A sampled answer is one observation, not stable coverage. Report it that way.
What to do with the results
Wrong facts in answers can trace back to unclear or inconsistent pages: a price stated two ways, an old service still described somewhere, a page that never says plainly what you do. Those are worth checking first because they are fixable. Being absent has many possible causes, and how much the rest of the web says about you is often one of them; it is not something your own pages alone can settle. The AEO vs SEO vs GEO guide covers what is worth changing on the page, and Wilde Digital’s AI search optimization scope ($1,500–$3,000 USD per scoped project) covers doing it, without any promise of citations.
Common questions
- Can I see exactly which questions led an assistant to cite me?
Partly. Bing’s AI Performance report shows sampled grounding queries for Microsoft’s AI experiences. Google’s report shows AI-feature impressions within Search Console. No platform gives a complete list of every question that produced a mention.
- Are AI visibility tracking tools accurate?
They run prompts at scale and report what they saw, which is a larger sample of the manual method above. Treat their figures as samples with their own method and limits, and check the method before trusting a share-of-voice number.
- Why does the answer change every time I ask?
Generated answers vary between sessions, users, locations and model updates. That is why a single check proves little and a repeated set of the same questions is more useful.
- Will blocking AI crawlers stop assistants mentioning me?
It can reduce it. Assistants that search the web rely on their search crawlers being allowed. Search crawlers and training crawlers are separate settings for several providers, so you can allow one and decline the other; decide deliberately and record the decision.
Want this applied to your business?
Send the site and the problem. You get an honest read on what is worth doing first, whether or not it is something Wilde Digital sells.