The Generative AI Report in Search Console: Do You Show Up in AI Answers?

What the new Generative AI performance report in Google Search Console shows, what it leaves out, and how to read a report that has no click data.

2026-08-11

Your customers no longer type every question into a search box and scan ten blue links. Some of them read the AI answer and close the tab. So — do you appear inside those answers? Until now there was no measurable way to tell. On June 3, 2026, Google announced Generative AI performance reports for Search Console. In this post we walk through what the report shows, what it does not show, and what you can actually do with the data you get.

Illustration of the Generative AI tab in the Google Search Console dashboard with an impressions chart for AI answers

This post is part of the AI Visibility Guide; see the guide for all four layers.

Where the Tab Sits

Search Console’s left-hand menu has a “Performance” section. Until now you would find essentially one report there: classic search results. In the new layout, the section listing all reports carries a separate performance report for Generative AI. There is a matching separate Generative AI report for Discover as well.

Keeping it separate is a deliberate design choice. Google does not blend these impressions into your classic search numbers; it presents them as two distinct measurements. Your impression count on the “Search results” tab and the number on the Generative AI tab should be read independently, not added together.

The report opened first, in June 2026, to a group of site owners in the United Kingdom. The reason was regulatory rather than technical: the control rights the UK Competition and Markets Authority (CMA) granted publishers. At the time Google said the reports would roll out more widely over time. As of August 2026 they have started appearing in accounts in Türkiye too — it is worth opening your dashboard after reading this, because the rollout does not reach everyone at once.

What Data Is in the Report?

The report has a single metric: impressions — how many times links to your site were shown to a user in generative AI features on Google Search. In the wording of Google’s official help page, “how many times links to your site were shown to a user in a generative AI feature.”

You can break those impressions down along four dimensions:

  • Pages: The final URL the generative AI feature linked to, after any redirects.
  • Countries: The country the search originated from.
  • Devices: Desktop, tablet, mobile.
  • Dates: By day, week, or month. Dates are calculated in the Pacific time zone.

Which features are counted? AI Overviews (the AI summary above search results) and AI Mode (the conversational search mode). Experiments in Search Labs are not counted; Google leaves them out because they are still under active development.

Two details keep you from misreading the report. First: if two links from your site appeared together in a single AI answer, that counts as one impression in the chart total. So summing the page-level numbers will never match the chart total. Second: the newest data is preliminary and may change within a few hours. The 1,000-row limit from the classic performance reports applies here too.

What It Does Not Show

The report’s most important characteristic is what is missing: there is no click data. Impressions only. There is no query breakdown either, so you cannot calculate a click-through rate; and there is no ranking indicator such as average position.

In the announcement Google said it is continuing to work with site owners to understand which insights would be most helpful, and that more metrics will be added over time. But as of August 2026 the picture is this: you can measure visibility, and you cannot measure what that visibility is worth in traffic.

What Does the Absence of Click Data Mean?

In short: visibility is not traffic.

In classic search results there is a learned relationship between impressions and clicks. If a page has 10,000 impressions and a 3% click-through rate, you can roughly estimate what doubling impressions would bring. On the generative AI side that bridge does not exist yet. Being named in an AI answer sometimes ends with the user visiting your site and sometimes with them reading the answer and closing it — and the report does not separate the two.

Three practical conclusions follow:

  1. Do not add these numbers to your classic impressions. Merging the two into a single “total search visibility” figure means reporting growth that did not really happen.
  2. Do not present this report as a revenue metric. Telling management “our AI visibility is up 40%” is not a measured business outcome as long as there are no clicks underneath it.
  3. Even so, it beats knowing nothing. This area used to be completely dark; now you can at least see the direction.

What Can You Do With This Data?

Single metric or not, the report makes you ask a useful question: which of my pages appear in AI answers, and what do those pages have in common?

A workable order of operations looks like this:

Export the Pages breakdown and put it next to the classic search report. Three groups will emerge: pages strong in both, pages strong only in classic search, and pages that show up only on the AI side. That third group is the most instructive — usually they are pages that answer a direct question clearly, with definitions and steps.

Write down the shared traits. Does the heading match a question? Is the answer given clearly in the first paragraph? Are there tables, lists, step-by-step flows? Is structured data marked up? These are practices we already know from the technical SEO side; the report makes it measurable, for the first time, whether they also work on the AI side.

Read the country and device breakdowns. If AI impressions from abroad are higher than you expected, you have a concrete case for investing in content in that language.

Look at the date chart after you change something. A single-metric report is still enough for comparative reading: the four-week curve after you rewrite a group of pages tells you something.

Over the past year this kind of work has been labelled generative engine optimization or answer engine optimization. Until now those headings rested largely on guesswork; this Search Console report replaces the guess with data specific to your own account.

There Is Also an Off Switch

In the same announcement Google gave site owners the ability to choose whether they appear in generative AI features at all. As Search Engine Land reported, this control also opened in the UK first, and Google stated that it is not used as a ranking signal for standard search results. If you opt out, your pages keep appearing in classic results; you only lose visibility in the AI features.

Do not treat this as a decision to rush. Opting out while you still have no click data means shutting down a channel you have not measured.

SEO Is Not Ending — a Layer Is Being Added

Reading this report as “classic SEO is over now” would be wrong. The same content and the same technical foundation feed both sides; crawlability, speed, clear headings, and correct markup do the same job in both places. What has changed is that visibility is now measured in two separate places, and those two measurements behave differently.

The AI side does not look like it will stop at visibility either: an arrangement is taking shape in which agents buy directly from APIs. We covered that side in detail in our post on preparing to sell to AI agents.

If there is infrastructure work to be done so that your site shows up properly on both of these layers, take a look at our technical SEO service or write to us directly through our contact page.