Google Search Console's Generative AI performance report tells you how often your pages showed up in AI Overviews and AI Mode. It does not tell you how many people clicked. Read it as a visibility signal, not a traffic number, and your 2027 planning will hold up.
The short version
- Google finished rolling the Generative AI performance report out to every site worldwide on August 31, 2026, so the data is already waiting in your Search Console whether or not you have opened the report.
- The report shows impressions only. There are no clicks, no click-through rate, and no average position, so you cannot derive AI-driven traffic from it.
- It covers Google's own surfaces: AI Overviews, AI Mode, and generative features in Discover. It does not cover ChatGPT, Perplexity, the Gemini app, or Claude.
- You can break impressions down by page, country, and device, but not by query, so keyword-level AI visibility stays a blind spot.
- A separate property-level control can remove your content from those AI features entirely, and it is not the same switch as Google-Extended.
What the report actually measures
The report counts one thing, and Google defines it narrowly. Per Google's Search Console documentation, an impression is logged when “links to your site were shown to a user in a generative AI feature.” That is a visibility count, not an engagement count. If your page is cited in an AI Overview that someone reads and never clicks, it still counts.
You get a useful set of cuts on that single metric. You can group impressions by page, which Google assigns to the canonical URL after redirects, by country, by device across desktop, tablet, and mobile, and by date in Pacific Time. There is a filter at the top of the report for text-based versus multimodal searches. What you do not get is a query dimension. The report will not tell you which prompts or searches surfaced your page, which is the gap most teams will feel first.
The page breakdown is the one genuinely useful cut here. It is the same page-level scoring logic I lean on when I score pages for a domain migration: find the handful of URLs doing the heavy lifting, then decide where attention goes.
Impressions without clicks is a reporting trap
The most common mistake I expect to see this quarter is someone dropping the AI impression number into a board deck or a quarterly business review and labeling it “AI traffic.” It is not traffic. It is the number of times a link to your site was displayed inside an AI answer, and a large share of those displays end with the user getting what they needed without clicking anything.
Google shipped the report without a clicks column, and that was a deliberate call it has said little about. The Search Central announcement frames the report around visibility, and independent analyses of the rollout confirmed there is no click data and no average position in it. So the honest read is this: impressions tell you that Google's AI features consider your content relevant enough to surface. They tell you nothing about whether that surfacing sent anyone to your site, or whether it quietly answered the question and kept the user in Google.
When I set the report up for a B2B software client last month, the impression curve looked great, up and to the right since September. The sessions from Google over the same window were flat. Both things were true at once. If we had reported impressions as traffic, we would have claimed a win that the pipeline numbers would have contradicted one quarter later. Visibility went up. Clicks did not follow. That is a strategy conversation, not a victory lap.
The measurement view that actually answers the question
To answer “is AI search sending us anything,” you need two reports in two different tools, because no single report spans both the impression side and the click side.
Search Console's Generative AI performance report is your impression and visibility layer, and only for Google's surfaces. For the click side, you are back in Google Analytics 4 (GA4), reading referral traffic, and only for the AI products that still pass a referrer. That caveat matters more every month, because several AI tools have stopped tagging their outbound links cleanly, which pushes those clicks into direct or unassigned traffic.
So the practical stack is simple. Use the Search Console report to see where you are visible across Google's AI features. Use GA4 referral and landing-page data to see what actually converts. Keep the two labeled separately in every report you hand upward: impressions in one column, sessions in another, and never a single blended “AI traffic” figure that implies a precision no tool currently gives you.
The opt-out control you should decide on before year-end
Sitting in Search Console's settings is a property-level switch that removes your content from AI Overviews, AI Mode, and Discover's generative features, and most teams have never made a deliberate call on it. Google's description is plain: opt out and you will receive no impressions or traffic from those AI surfaces, and the setting is not used as a ranking signal anywhere else in Search.
Here is the part that trips people up. This control is not Google-Extended. Google-Extended is the robots.txt token that governs whether your content is used to train Google's Gemini models, and toggling it has no effect on whether you appear in AI Overviews. Two different levers, two different questions. One is whether you want to show up in Google's AI answers. The other is whether you want your content training Google's models. Decide them separately.
For almost every B2B SaaS site, the answer on the AI-features control is leave it on. Visibility in an AI Overview is still visibility, and opting out to protect clicks you cannot measure is trading a known upside for a feeling. The teams that should think hardest about opting out are the ones whose content is the product, like gated research or proprietary data, where being summarized for free is a real loss. That is a small minority.
What this does not cover
This report is deliberately narrow, and a few things sit outside it. It does not measure any non-Google AI surface, so your presence in ChatGPT, Perplexity, or Claude answers is invisible here and needs separate tooling. It does not attribute clicks or revenue, so you cannot connect an impression to a signup. It does not expose queries, so you cannot build a keyword strategy from it the way you can with the classic Performance report. And it cannot tell you whether an impression influenced a buyer who later arrived through another channel, which is the question your revenue team will actually ask. Those are real limits, not bugs, and pretending the report answers them is how a measurement program loses credibility.
FAQ
Does Search Console's AI report show how much traffic AI search sends my site?
No. The Generative AI performance report shows impressions only, meaning how often your links appeared in a generative AI feature. It has no clicks, no click-through rate, and no position data, so you cannot calculate AI-driven traffic from it. For clicks, you need GA4 referral data, and even that is incomplete because some AI tools strip their referrer.
How is the generative AI control different from Google-Extended?
They control different things. The Search Console setting decides whether your content can appear in AI Overviews, AI Mode, and Discover's generative features. Google-Extended, a robots.txt token, decides whether your content is used to train Google's Gemini models. Opting out of one has no effect on the other.
Can I see which keywords trigger my pages in AI Overviews?
No. The report has no query dimension. You can break impressions down by page, country, device, and date, but not by the search or prompt that surfaced your page. Keyword-level AI visibility is still a gap.
When did the report become available to everyone?
Google rolled it out to all websites worldwide on August 31, 2026, after a phased start in June. If you have a Search Console property, the report and the historical data behind it are already there.
Should a B2B SaaS company opt out of AI features?
Almost never. For most sites, appearing in AI answers is visibility worth keeping, and opting out trades a known upside for clicks you cannot currently measure. The exception is sites whose content is the product, such as gated research, where free summarization is a genuine loss.
If you want me to stand up this two-tool measurement view for your site, map the pages that are actually visible in Google's AI features, and give your exec team a number that will survive the next quarter, book a 20-minute Search Console read-out.