Google explains AI impression counting in Search Console

Search Engine Journal reports that Google's John Mueller has clarified exactly how impressions are counted in Search Console's AI search report, and the mechanics matter more than the headline number suggests.

The core distinction is this: Search Console logs an impression when a link to your page appears in an AI-generated response. But not all links are equal. Some appear immediately, visible to any user who sees the answer. Others are "user-activated," collapsed behind an interface element the user must click to expand. Those only register an impression after a user actually activates them.

This is not a minor bookkeeping note. It means two pages can show identical impression counts while having fundamentally different visibility profiles. A page cited in a prominent, immediately visible link has been seen by everyone who received that AI answer. A page buried in a collapsed attribution panel has been seen by nobody until someone chooses to look, and many never do.

What the numbers are actually measuring

The practical consequence for brands is that raw impression data from the AI report is, on its own, ambiguous. High impression counts could reflect genuine prominence in AI-generated answers, or they could reflect citations that users never encounter unless they actively dig for them. Mueller's clarification does not resolve this ambiguity; it defines it.

For a multilateral institution publishing policy guidance, or a financial services firm whose compliance documentation surfaces in AI answers, the gap between "cited" and "seen" is operationally significant. If a UN agency's report is referenced only in a collapsed source list that fewer than 5% of users ever expand (a reasonable inference from comparable UI research on accordion-style elements), then the impression figure in Search Console flatters the actual reach considerably.

The mechanism also raises a calibration problem for anyone benchmarking AI search performance. If impression counting methodology changes as Google refines its AI interfaces, historical trend lines break. A spike in impressions might reflect increased citation frequency, or it might reflect a product decision to surface more links in an expanded default view. Mueller's explanation does not indicate whether Search Console will label which impressions came from which link type, which would be the minimum required for the data to be analytically clean.

Google has not announced that it will break down impressions by link type in reporting. Until it does, the AI search impression metric functions more as a floor than a ceiling: it tells you that your page was referenced somewhere in the answer environment, but not whether any human eye landed on the link.

For B2B brands whose authority depends on being visibly cited, rather than merely technically present, the implication is direct. Optimising for AI citation frequency is necessary but insufficient. The next question is where within the AI answer your citation appears, and that information is not currently in Search Console.

Teams tracking AI visibility should treat the current AI impression report as an existence check, useful for confirming that pages are entering the citation pool at all, not as a reliable measure of brand exposure. The click data, where it exists, will be more diagnostic. A high impression count paired with near-zero clicks almost certainly indicates a user-activated link that most users never expand.

Mueller's clarification is useful precisely because it makes the limitation legible. Brands that treat AI impressions as equivalent to traditional organic impressions will systematically overestimate how much of their content users are actually encountering in AI answers.