Last updated September 2026.
On Aug. 15, 2026, Search Engine Land reported that Google keeps widening access to the Generative AI Performance report, the panel inside Search Console that shows how a site’s pages surface inside AI Overviews and AI Mode. Google’s own Search Central Blog confirms the same expansion. The report first launched to UK sites on June 3, 2026. More than two months later, it still ships without a single click number.
That gap matters more than a normal missing metric would. This report is currently the only first-party data Google gives site owners for AI answer surfaces. No other feed tells you, straight from Google, how often your page fed an AI Overview or an AI Mode answer. It is the closest thing the GEO industry has to an official measurement standard, but a standard built on impressions alone cannot tell you whether any of that visibility turned into a visitor.
The same rollout added a new control: a toggle that lets a site block its content from AI features entirely. Turning it on trades away your only first-party visibility signal for those surfaces, along with whatever traffic the feature might send. Most sites should leave it off and use the data instead of hiding from it.
For a marketer trying to justify a GEO budget line, this report is the one number a skeptical stakeholder cannot dismiss as a vendor’s estimate. It comes straight from the platform that owns the surface. That credibility is exactly why the missing click column stings. A report a stakeholder trusts, that still cannot answer “did this drive a visit,” pushes the burden of proof onto whatever you pair it with.
What the report shows vs. what it withholds
Compare the two columns before you build a measurement plan around this report.
| Search Console shows you | Search Console withholds |
|---|---|
| Impressions by individual page | Click counts |
| Impressions by country | Click-through rate |
| Impressions by device | Sessions or on-site behavior after the click |
| Impressions by date | Which exact AI answer or prompt generated the impression |
| A toggle to block AI features on your content | Revenue or conversion signals |
Every row on the left is real, first-party data straight from Google. Every row on the right is a blank you have to fill in yourself.
Why impressions alone can’t prove the traffic case
An impression tells you Google’s AI features treated your page as a fit for an answer. It does not tell you whether a person read that answer, clicked through, or ever heard of your brand as a result.
Zero-click behavior is already common on standard AI Overviews. A chat-style AI Mode answer can go further still, since the model can summarize your point without sending the reader anywhere at all. Impressions can climb for months while your actual referral traffic sits flat, and Search Console alone gives you no way to tell those two stories apart.
That distinction is the whole measurement problem in one sentence. Impressions describe reach. Clicks describe outcome. A report that only ever shows reach cannot settle a budget argument on its own, no matter how many pages, countries, or dates you slice it by.
Triangulate in 3 steps
Google is not going to hand you clicks. Until it does, build your own measurement stack around three moves.
- Pull your AI impressions baseline from Search Console. Filter the Generative AI Performance report by page, country, and device, and log it on a recurring schedule so you see a trend, not just one snapshot.
- Add a third-party citation tracker that reads live AI answers. Profound, Temso, Peec AI, and Ahrefs Brand Radar each run prompt panels against live chat interfaces and log exactly which prompts cited your domain, a layer Search Console does not expose.
- Cross-check both against a downstream signal you already own. Branded search volume, direct traffic, and referral hits from known AI domains in your existing analytics will not prove a single citation drove a single visit. But a real move in impressions and citations that lines up with a real move in one of those signals is stronger evidence than any one source alone.
None of those four platforms can see your Search Console impressions data. That number stays inside Google’s own dashboard, tied to your verified property. What they add is the citation-level detail Search Console was never built to show: the exact prompt, the exact answer, and whether you were named and linked or just paraphrased without credit.
Run the two side by side for a full reporting cycle before you draw a conclusion. A page with rising Search Console impressions and rising third-party citations is gaining ground on AI surfaces, even before a single click number exists to prove it. A page with rising impressions and flat citations is worth a closer look at what the AI answer is actually doing with your content.
Full profiles for all four tools, plus five more trackers, sit on the nine-tool LLM visibility benchmark. The measurement protocol page documents how this site scores tracking accuracy and model coverage across that comparison, the same two questions worth asking about any tool you use to fill this specific gap.
Check your own gap this week
Open the Generative AI Performance report in your own Search Console property and read the impression trend for your top pages. Then open the nine-tool LLM visibility benchmark and pick a citation tracker that shows you what Google’s report leaves out.