Research
Google Search Console Finally Has an AI Performance Report. Here Is What It Shows and What to Do With It.
The generative AI report finished rolling out worldwide on August 31. Most coverage stops at what it shows. The useful part is what it hides, and the workflow that gets value out of impression-only data anyway.
For two years, SEOs asked the same question: how much of my visibility is happening inside AI Overviews and AI Mode? Until this summer, Google’s answer was a shrug. AI impressions were folded invisibly into the regular performance report, and you could not separate them.
That changed. On June 3, 2026, Google announced dedicated generative AI performance reports in Search Console, and by August 31 the rollout reached every property worldwide. If you own a site, the report is sitting in your Search Console right now.
This post covers what the report measures, the three things it still refuses to tell you, and a weekly routine for using it anyway.
What the AI performance report shows
The report covers two surfaces:
Generative AI performance (Search): your impressions inside AI Overviews and AI Mode on Google Search.
Generative AI performance (Discover): your impressions inside AI features in Google Discover.
Within each, four familiar dimensions:
| Dimension | Detail |
|---|---|
| Page | Which URLs earned impressions inside AI features |
| Country | Where those impressions happened |
| Device | Desktop, mobile, tablet (Search report) |
| Date | From hourly granularity up to monthly views |
The page-level breakdown is the useful one. It tells you which of your pages Google’s AI systems consider quotable. That is a different bar, and arguably a more demanding one, than which pages rank. A page can sit at position 8 in classic results and never appear in an AI Overview, or appear in AI Overviews for queries where it does not crack the classic top 10 at all.
The hourly granularity matters more than it looks. AI Overview presence is volatile. Google adds and removes the feature per-query constantly, and an hourly view lets you catch when a query you care about gains or loses its AI Overview. A daily rollup smooths that away.
How impressions are counted
This matters more than it sounds, because the counting rules differ from the standard report:
- An impression is counted each time a link to your site appeared in a generative AI feature, whether or not the user saw or clicked it
- If two URLs from your site appeared in the same AI answer, that counts as one impression in the chart total, not two
- Apply a URL filter and the chart switches to counting per-URL rather than per-property
- The most recent data may show as a dotted line, meaning it is still being collected and can update within hours
The one-impression-per-site rule changes how you read big numbers. A property with very high AI impression counts likely has multiple pages being sampled across individual answers, and the per-page view understates the property total.
The three things it hides
The report shows impressions only. Three data points are missing, and each absence changes how you should read the numbers.
1. No clicks or CTR
You can see that your page appeared in AI features 5,000 times. You cannot see whether a single person clicked through. Google has not committed to a timeline for adding click data.
This is the report’s biggest limitation, because impressions inside an AI answer are not like impressions in classic results. In an AI Overview, your page may be one of several collapsed citations behind the answer. Visibility, with no realistic chance of a visit.
Treat AI impressions as a citation metric, not a traffic metric. We cover the estimation workarounds in how to estimate AI Mode clicks when GSC won’t tell you and the broader measurement stack in measuring AI visibility.
2. No queries
The regular performance report tells you which searches triggered your impressions. The generative AI report does not. You know which page appeared in AI features, but not for what question.
That gap kills the most natural optimization loop: find queries where you are cited, strengthen those pages, find near-miss queries, close the gap. For now, the closest substitute is cross-referencing. Take your top AI-impression pages, pull their query lists from the regular report, and assume overlap. Imprecise, but it is the only move available.
Google has said it is working with publishers on which additional metrics would be most useful, so the query omission looks like a privacy and engineering problem rather than a permanent decision. Plan for the current report, not the promised one.
3. No clean split between AI surfaces
AI Overviews and AI Mode are different surfaces with very different behavior, and blended totals muddy the signal. The report structure has already shifted once since the June launch, so watch how Google evolves the breakdowns.
The first comparison worth running
The single most useful thing to do with this report on day one: put your top pages by classic impressions next to your top pages by AI impressions.
They will not be the same list.
Google’s AI systems and Google’s classic ranker make different choices about the same content. A page can rank top 10 in blue links and never get cited, while a mid-ranking page becomes a regular AI citation because one section answers a question cleanly. That delta is the report’s value right now. It is the first first-party evidence of what Google’s generative systems actually trust on your domain, page by page.
Expect the familiar informational-over-commercial skew too. Blog content gets cited in AI features far more than service pages. If your commercial pages show near-zero AI impressions, that is normal, not a defect.
The blocking control
Alongside the reports, Google added something with more strategic weight: a control that lets site owners block their content from AI features without leaving classic search. AI Overviews, AI Mode, and generative AI in Discover, switched off for your site if you choose.
The origin story matters. Google introduced the control after a UK regulatory order requiring more publisher control over content used in AI features, with UK site owners getting it first from June 17, 2026 before wider rollout.
For most sites, blocking is the wrong call. Pages cited in AI Overviews earn measurably more organic clicks than uncited competitors, and opting out surrenders the surface to whoever stays in. Subscription publishers and some YMYL sites have a genuine case. We wrote the full decision framework in how to block your site from Google’s AI features, and whether you should.
A weekly routine that uses the report properly
Fifteen minutes, once a week:
- Log total AI impressions for the property. One number, one row in a sheet. You are building the trendline Google will not show you retroactively.
- Pull top 10 pages by AI impressions. Flag anything that entered or left the list versus last week.
- Cross-reference each new entrant against the regular report’s queries for that page. Note the likely triggering queries.
- Check the ratio of AI impressions to classic impressions. A rising ratio with flat clicks is the signature of zero-click drift. That is your early warning to strengthen presence inside the answer, through citations and branding, rather than chasing clicks it can no longer send.
- Feed page-level winners back into your content plan. Whatever structure your AI-cited pages share, do more of it. In our own classic-search data, the pages that earn clicks are consistently data-first posts with early, direct answers, the same pattern the research on AI visibility KPIs keeps pointing to.
Pair this with GA4 referral data for the click side of the story. Our GA4 for AI SEO tracking guide covers that setup.
Next in this series: our own recovery data. What a Google quality hit actually looks like in Search Console, charted from this site’s export.
