Build1 publisher2 min readPublished
Google's Search Console counts AI Overview and AI Mode impressions by canonical page
Google Search Console now counts page impressions in AI Overviews and AI Mode, in reports that reached all websites worldwide by August 31, 2026. Teams can confirm their pages show up in AI answers, but linking those appearances to visits takes other data.
The Engineer · Build desk
What happened
- Google announced the reports on June 3, 2026, and first tested them on a subset of websites, initially in the UK.
- Impression counts can be broken out by page, country, device and date.
- The reports do not show which queries triggered an AI appearance, and they include no click-level data.
- Each impression is tied to the final page URL and rolled up under its canonical URL.
- The reporting also covers generative AI features in Discover, beyond AI Overviews and AI Mode in Search.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- constraint Content teams cannot see which questions put their pages into AI answers, so topic decisions have to be inferred from which canonical pages gain impressions.
- decision Sites with heavy canonicalisation or redirects have to map which variants roll up to which canonical before reading any page row as one URL's visibility.
- cost Any claim that AI answers drive visits or revenue still has to be assembled by hand from the standard performance report and on-site analytics.
- capability Teams can now track AI-feature visibility as its own series inside Search Console and compare it by country and device.
Of the four dimensions, the page key needs the most care [6]. Impressions attach to the final URL, so an address that redirects counts against wherever it lands [7]. A cluster of near-duplicate pages whose canonical tags point at one URL shows up as a single row [7]. According to the dev.to write-up of Google's announcement, teams running similar pages, redirects or canonical tags should account for this before assuming a row stands for every URL variant [10]. For a catalogue with parameterised or regional duplicates, the page view cannot say which variant Google actually showed [7].
Without queries, the closest thing to demand data is watching which canonical pages gain or lose impressions over time [5]. The write-up's recommended process starts there. Track trends by date, then review which canonical pages are receiving impressions [11].
Someone will paste this chart onto a slide titled AI traffic. For impressions to stand in for traffic, two things would have to be true. Appearances in AI features would need a steady click-through rate, and that rate would need to be similar across the pages being compared. This report cannot test either assumption. The write-up says a rising count cannot by itself establish whether appearances produced clicks, leads or sales [12]. Attribution still means lining the report up against the standard performance report and on-site analytics [11]. Matched on page and date, that join can show a correlation at best.
The design choice I would defend is the separation. Before this rollout, generative AI appearances had no reporting path of their own in Search Console [9]. Now they are counted in dedicated reports, apart from the rest of Search Console reporting [1]. Some content teams mostly need to know whether their explainer pages get surfaced at all. For them, I think a first release that only counts visibility is the right tradeoff. I would use it to check presence over time, and keep revenue questions in analytics.
What to watch
- Whether Google adds query or click data to the Generative AI reports, the change that would let appearances be tied to visits.
- How AI-feature impressions move between page rows when a site changes its canonical tags or redirects.