Google GenAI Performance Reports: What the New GSC Data Offers

A few weeks ago I wrote here that Microsoft had sent a clear signal to Google with the AI Performance Report in the Bing Webmaster Tools. Bing showed dedicated AI citation data, while Google still bundled its AI Overviews numbers into the regular web report. The gap was clearly visible.

Google added that. On June 3, 2026, the Search Console team announced the “Search Generative AI performance reports”, a dedicated view of how visible your pages are inside Google’s AI answers. Since 31 August 2026 every property worldwide has it.

For me, the first question with announcements like this is always the same: what does the blog post actually say, and what did the trade media add around it? That’s exactly what I separate cleanly here. I’ll show you which data you really get, which metric is still missing and where the limit sits that bites next now that the rollout is done. It is no longer availability. It is access.

According to the official blog post by Hillel Maoz (Search Ecosystem Engineering Manager) and Moshe Samet (Product Manager Lead, Search Console), the new reports launch with “dedicated reports for Search and Discover, to help you understand your site’s visibility within generative AI features on Search”.

In plain terms: you get a dedicated view showing your pages’ visibility in generative AI features. One point that’s easy to miss: Google announces that it will additionally make this data available in the existing Performance report, so you retain the overall picture of your visibility. What’s new is the separate view on top.

Google started deliberately slowly. In its own words: “We are rolling these reports out to a subset of websites, allowing us to thoroughly test them and receive feedback before making them widely available.” That sentence was still there when I checked on 3 September 2026, but it has been overtaken. At the top of the same blog post Google added a note: “As of August 31, 2026, we’ve rolled out these insights to all websites worldwide.” Just under three months passed between announcement and completion.

This official screenshot already shows the key catch I’ll come back to in a moment: the header reads only “Total impressions” - there is no clicks column. The view also carries a “Beta” label, and the sidebar lists “Generative AI” under both Search results and Discover.

Google lists exactly which dimensions the reports show. I’ve summarized them here with what each one means:

The clicks not yet shown are the crux. Visibility is nice, but for a profitability calculation you need the click. Google phrases it cautiously: “We’re continuing to work with website owners to understand what insights and data would be most helpful to inform their strategies, such as adding additional metrics over time.” Translated: clicks may come, nothing is promised.

What Google says: Since the end of August the announcement carries a note at the top: “As of August 31, 2026, we’ve rolled out these insights to all websites worldwide.” No country is excluded. On the same day Google also enabled the control for excluding your content from AI features globally. Measuring and switching off arrived together, which is unlikely to be a coincidence: without numbers, the decision to switch off would be a bet.

Where to find it: In Search Console as an expandable section under the performance report, separated into Search results and Discover. If you do not see it, Google puts that down to data volume rather than the rollout. Small properties without meaningful AI impressions will find the view empty.

The limit that matters now: The report exists only in the interface. As of 3 September 2026 the Search Console API reference lists exactly six values for the type parameter: discover, googleNews, news, image, video and web. There is no value for the generative AI features, and the page does not mention generative AI anywhere. According to the documentation, web returns the combined “All” tab, where AI impressions are included but cannot be separated out.

In practice that means anyone who wants these numbers in a dashboard, in Looker Studio or in their own script cannot get at them as things stand. What remains is a manual export from the interface, per property. For a single site that is tedious. For anyone looking after twenty client accounts it is a real problem. That is the point where something should move next now that the rollout is done.

That was exactly the point from my Bing article: according to the official Bing Webmaster blog, Microsoft delivers citations and grounding queries - metrics for which there is simply no equivalent on Google’s side. The new reports change little about that for now. Google shows that you appear in AI answers (impressions). Bing additionally shows how your content is cited and drawn on to generate answers.

From my point of view this isn’t an either-or. Anyone serious about their AI visibility uses both data sources: Google for reach in the larger search ecosystem, Bing for the deeper citation mechanics. Now that the Google report is available everywhere, the argument for Bing has shifted. It is no longer about bridging a gap, but about the metrics Google still does not deliver.

Here’s my concrete plan you can act on right away:

1. Open the report and note your starting value: The section sits under the performance report, separated into Search results and Discover. Note the impressions of your most important URLs for today. A Search Console annotation records the date.
2. Plan the export: The Search Console API does not give you these numbers as things stand. Anyone reporting regularly should book the manual export as a fixed appointment rather than wait for automation.
3. Add Bing AI Performance: The Bing report delivers citations and grounding queries, which Google does not have. Setting it up takes ten minutes.
4. Surface AI-relevant questions: With the GSC regex for AIO user questions, you can find which question queries your audience asks - exactly the ones that trigger AI answers.
5. Increase citability: Clean entities help AI systems capture your content unambiguously. How much structured data really helps for AI Overviews I’ve broken down separately. The short version: Google requires no special schema. An Ahrefs study found no positive effect on citations.
6. Read impressions without clicks correctly: Interpret AI impressions as a visibility and brand signal, not a traffic promise. The missing click reference changes how you measure success.

This section was added on 21 August 2026. It belongs here because it concerns the reliability of exactly the report this article describes.

Google lists the incident on its data anomalies page and describes it there as follows: “A logging error caused a decrease in impressions on the Generative AI performance report in Search for data from August 13 - August 17, 2026. This issue affects data logging only.”

Two details in that are practically usable. The window is bounded on both sides, so five days from 13 to 17 August 2026. And the error sat in the recording, not in the visibility. Anyone who saw a drop in those days saw a counting error.

When I prepared this addendum, something else was standing there. On 19 August the same entry carried no end date but the addition “and is ongoing”. Between 19 and 21 August Google closed it. I found no report about that, even though the drop itself went through the trade press on 17 August. The end of an incident is apparently the less attractive half of the news.

A second entry of the same date concerns the Discover report. That one is cut differently: only 13 August 2026, but with clicks and impressions. For properties with Generative AI features in Discover the same error additionally lowered the reported impressions. Do not pull the two entries together, they describe different reports and different periods.

For how you rate this report that means nothing dramatic, but it does mean something. A time series that starts in June and has its first documented gap in August is young. That is not an accusation against Google but a statement about precision: for now, treat the curve as an indication of an order of magnitude, not as a measuring instrument. How to narrow down a drop systematically is in Impressions collapsed.

For me, the announcement shows above all one thing: AI visibility is gaining importance as its own reporting topic and needs its own numbers. That Google would follow Bing with a comparable report was foreseeable. That it stays with pure impressions for now and the rollout starts slowly also shows how early we still are in this development.

My advice stays sober: enjoy the new transparency, but don’t build your strategy on it alone. Right now the most reliable AI data comes from combining both worlds - and from content good enough to be cited in the first place.