AI Overviews took the clicks

Search Console has no AI Overviews report. Google's documentation on AI features says sites appearing in them "are included in the overall search traffic in Search Console," inside the Web search type, with no separate breakout and no filter to isolate them. The same page offers one claim in your favor, that "when people click from search results pages with AI Overviews, these clicks are higher quality (meaning, users are more likely to spend more time on the site)," but publishes no site-level data a publisher could verify. That is the honest state of the official record: the feature that moved your clicks is invisible in the tool that counts them.

The best independent published measurement is Pew Research Center's July 2025 study of real browsing data from 900 U.S. adults during March 2025. Users who got an AI summary clicked a traditional result on 8% of visits, against 15% when no summary appeared. Links cited inside the summary itself were clicked on just 1% of visits. And searchers ended their session entirely on 26% of pages with a summary, versus 16% without. One study, one month, one panel, so treat it as the scale of the effect rather than your number. But it is measured, published, and dated, which beats every invented percentage in a sales deck. It also quietly retires the hope that "being cited in the Overview" replaces the click. At 1%, it does not.

So the measurement is inference: you cannot filter for AI Overviews, but they leave a fingerprint your own data can show.

1. Set an honest comparison

Performance report, Web search type, comparing this quarter against the same quarter last year. Year over year matters here: comparing against last month confuses AI absorption with ordinary seasonality, and this shift built over quarters, not weeks.

2. Look for the three-metric signature

Impressions steady or up. Average position stable, within a point or so of where it was. Clicks and CTR down. That combination has almost no other cause: you kept the rankings and lost the visits. If position fell too, stop here, because you are looking at a ranking loss and the update diagnosis comes first.

3. Segment by query class

Export queries for both periods and sort them into rough classes: informational questions (how, what, why, can, does, versus), brand searches for you by name, and action queries where the searcher needs a tool, a current number, or a decision. Compute CTR change per class. The AI Overviews fingerprint concentrates in the informational class; brand queries and action queries usually hold. That contrast is the diagnosis. A site-wide uniform CTR drop across every class, brand included, points somewhere else, often at how your results are displayed rather than at AI answers.

4. Price it

Clicks lost per class times your revenue per visit gives the honest cost of the shift. That number, not a feeling, is what the pivot-or-persist decision should run on.

Your own class-by-class export will show you this on your site, and the pattern across publishers is consistent in direction: an AI answer absorbs the query when the answer is the whole job. It cannot absorb the visit when the visit is the job. Queries where clicks tend to hold: tools and calculators the searcher must operate, data that changes and must be current, decisions with real stakes where a summary is not trusted alone, first-person experience a synthesis cannot manufacture, and brand queries from people who wanted you specifically. Queries that tend to hollow out: definitions, generic how-tos, and any question fully answered in three sentences.

We do not publish a percentage for how much each class holds, because we have not measured one and Google does not publish one. The direction, though, is the strategy: the defensible share of your site is whatever a searcher still has to click for. Count what your holding classes have in common, and build the next quarter's content there rather than adding to the class that is being answered above you.