For most of the answer-engine optimisation (AEO) era, teams have measured Google’s AI surfaces by inference: scraping screenshots, buying third-party trackers, or squinting at referral traffic. Search Console’s generative AI performance report changes that. For the first time, Google itself tells you when links to your site were shown to a user inside a generative AI feature on Search. That is a real milestone for first-party measurement — and it comes with sharp, documented limits that determine what you can honestly conclude from it. This guide covers both.

What Google actually shipped

On 3 June 2026, Google announced Search Generative AI performance reports in Search Console — dedicated reports for Search and for Discover — giving site owners a separate view of their visibility within generative AI features. For Search, the report covers AI Overviews and AI Mode; Google’s help centre documentation states the list is expected to be updated over time, and that experiments in Search Labs are excluded.

Per the announcement, the report shows:

• Impressions — how often URLs from your site appeared in generative AI features in Search and Discover;
• Pages — which URLs appeared within AI features (attributed to the canonical URL, as in the standard report);
• Countries — visibility on a country basis;
• Devices — desktop, tablet or mobile (available for Search results);
• Dates — performance over time, with all dates in Pacific Time.

One detail that matters for anyone reconciling numbers: this data is included in the overall performance report’s Web search type, not added on top of it. The new report is a dedicated lens on impressions you were already accruing, so never sum the two views — you would be double counting.

What it shows versus what it withholds

The cleanest way to understand the report is side by side with the standard Search performance report it descends from.

Each absence is worth spelling out. There are no clicks or CTR — Google’s blog post notes it is working with site owners on “adding additional metrics over time”, but today you cannot see whether an AI Overview appearance sent anyone to your site. There is no query dimension, so you cannot tie impressions to the prompts or questions that triggered them. And there is no API route: Google’s Search Analytics API documentation lists only discover, googleNews, news, image, video and web as search types, with no generative AI value — so automated pipelines are limited to whatever you download via the report’s export button.

Where the rollout stands — and an early data wobble

Google launched the report to a subset of properties for testing, and as of late August 2026 its help documentation still carries that subset-rollout note. On 11 August 2026, Search Engine Roundtable reported the report appearing across many more properties — initially calling it fully live, then updating the piece to say it was not fully live yet but had expanded to many, many more profiles. Google has not formally announced a completed rollout, so if you cannot see the report, you may simply not have it yet — or your site may not have enough generative AI impressions to populate it.

Trust the trend line carefully, too. Google’s own data anomalies page records that a logging error caused a decrease in impressions on the generative AI performance report in Search for data from 13–17 August 2026, affecting data logging only. If your chart shows a mid-August dip, that is probably Google’s plumbing, not your content.

Apply For Partnership

What a SaaS team can genuinely learn from it

Within those limits, the report supports several honest workflows:

1. Baseline and trend your Google AEO visibility

You finally have a first-party impressions series for AI Overviews and AI Mode combined. Export it weekly, chart it against your total Web impressions, and you get a defensible “share of our Google visibility that is generative” ratio — a far better input for internal reporting than screenshots. If you already track a share-of-model metric across answer engines, this becomes the Google-surface component of that scorecard, measured by the engine itself rather than by sampling.

2. Find which pages answer engines actually surface

The pages dimension shows which canonical URLs earn generative AI impressions. In our experience the winners are rarely the pages you would guess: definitional posts, FAQ-style pages and comparison content tend to surface, while product pages lag. That tells you where to concentrate answer-capsule formatting, schema and refresh effort.

3. Segment by country and device

For a SaaS business selling across markets, the country split shows whether your AI-surface visibility is concentrated where your buyers are. Device data (for Search results) tells you whether AI Mode exposure skews mobile — relevant if your signup flow underperforms on small screens.

4. Run before-and-after content experiments

Because the data is page-level and date-stamped, you can restructure a post and watch whether its generative AI impressions move over the following weeks. Without a query dimension you cannot prove which question you won, but a page-level lift after a targeted rewrite is still usable evidence — the same standard of honest inference we argued for when reading 2026 AEO benchmark reports.

What it cannot tell you

Four limits keep the report from being a complete AEO measurement system:

• Visibility is not traffic. Impressions without clicks can flatter you. Pew Research Center’s analysis of March 2025 browsing data found users who encountered an AI summary clicked a traditional result link in 8% of visits, versus 15% when no summary appeared — so a rising impression line can coincide with flat or falling sessions. You need downstream measurement to know.
• No queries means no intent. You cannot see the prompts behind your impressions, so you cannot separate “surfaced for our category” from “surfaced for something incidental”. Page-level proxies are the best you get.
• It is Google-only. ChatGPT, Perplexity, Copilot and Claude are invisible here. A Google-surface impressions chart says nothing about your standing in the engines a growing share of buyers actually ask.
• Young data, moving pipeline. A logging error within the report’s first weeks of wide availability (13–17 August 2026), preliminary-data dotted lines, Pacific Time bucketing and property-versus-URL aggregation quirks all mean you should annotate charts and avoid over-reading short windows.

One adjacent control is worth knowing: alongside the report, Search Console is rolling out a Search generative AI control that lets a site exclude its links and content from Search generative AI features entirely. Inclusion is the default, and Google states the control is not used as a ranking signal elsewhere in Search. For most SaaS teams optimising to be cited, the relevant point is simply that exclusion exists — and that an excluded site will record no impressions here at all.

How it complements GA4-side measurement

The report and your analytics stack answer different halves of one question. Search Console tells you how often Google’s AI surfaces showed you; GA4 tells you which AI surfaces — Google’s and everyone else’s — actually sent you people, and what those sessions did next. We’ve covered the GA4 side in detail in our guide to measuring AI referral traffic in GA4, so here we’ll just note the join: reconcile the two at the page level. A page with strong generative AI impressions and weak AI-attributable sessions is being summarised without being visited — a candidate for stronger differentiated content that a summary can’t replace. A page with weak impressions but real AI referrals is likely earning citations outside Google’s surfaces. Both patterns change what you write next.

This upstream-downstream pairing matters even more if your funnel is built to convert the traffic AI does send. At Zian AI we see this from the response side: an answer engine can surface you at any hour, and the visitor it sends expects an immediate, competent reply. That’s the gap our digital team agents — from 24/7 multilingual customer support to outbound appointment setting — are built to close while you sleep.

Sources & ownership

Frequently asked questions

Does the generative AI performance report show clicks from AI Overviews or AI Mode?

No. The report currently shows impressions only, segmented by page, country, device and date. There is no click, CTR, average position or query data, though Google’s launch post says it is considering adding additional metrics over time based on site-owner feedback.

Do AI Overviews actually reduce clicks to websites?

Independent evidence says yes. A Pew Research Center analysis of March 2025 browsing data from 900 US adults found users clicked a traditional result link in 8% of visits where an AI summary appeared, versus 15% of visits without one, and clicked a link inside the summary itself in just 1% of those visits.

Can I pull the generative AI data through the Search Console API?

Not at present. The Search Analytics API’s documented search types are discover, googleNews, news, image, video and web, with no generative AI option. The report itself offers an export button for downloading chart and table data, so automation currently means scheduled manual exports.

Are generative AI impressions added on top of my normal Search impressions?

No. Google states the generative AI report draws on data already included in the Web search type of the overall performance report, so the new report is a filtered view rather than an additional bucket. Adding the two together would double count.

Why does my report show an impressions dip in mid-August 2026?

Google’s data anomalies page records a logging error that caused a decrease in impressions on the generative AI performance report in Search for data from 13 to 17 August 2026, affecting data logging only — actual visibility was not necessarily down.

Why can’t I see the report in my Search Console property?

Two documented reasons: Google is still rolling the report out and not all properties have access yet, or your site has not received enough impressions in generative AI features on Google Search — including if it has been excluded from those features via the Search generative AI control.

First-party visibility data is necessary for AEO, but it only measures the top of the funnel. If you want the enquiries that AI surfaces do send to be answered, qualified and booked automatically — in 30+ languages, around the clock — join the Zian AI waitlist beta.

Apply For Partnership