Google Search Console Now Reports Generative AI Impressions. Here's What SEOs Can Actually Measure.
Google added dedicated Generative AI performance reports to Search Console for AI Overviews, AI Mode, and Discover. Here's what the report measures, what it still misses, and how SEOs should use it.
Google has finally given SEOs an official AI visibility report inside Search Console. That is the big news from the June 3, 2026 launch of Search Generative AI performance reports. The catch is just as important: this is an impressions report, not a complete AI attribution system.
That distinction matters. If you read the new report as a way to see which URLs are appearing inside AI Overviews, AI Mode, and generative AI features in Discover, it is useful right away. If you expect it to explain prompts, clicks, conversions, citation position, or why Google picked one page over another, you will overread the data.
The launch was announced in the Google Search Central Blog by Hillel Maoz and Moshe Samet. Google is rolling the reports out to a subset of websites first so it can test them and gather feedback before wider availability.
What Google Actually Launched
Google launched dedicated Generative AI performance reports for Search and Discover. The Search version covers AI Overviews and AI Mode. The Discover version covers generative AI features in Discover.
The data still rolls into the regular Search Console performance reporting system. What changed is the separate view. Instead of trying to infer whether an impression came from a standard result or an AI feature, eligible properties can now inspect the AI slice directly.
The Search Console Help documentation adds a few important boundaries. Search Labs experiment data is excluded because those experiments are still in active development. The report includes data from the Web search type. Regular Search performance limits apply, including the 1,000-row table limit.
If you do not see the report yet, that does not prove your site has no AI visibility. Google lists three likely reasons: your property may not be included in the rollout, the site may not have enough impressions in supported generative AI features, or the site may have excluded itself from Search generative AI features.
The Single Metric You Get Today
The current core metric is impressions.
Google defines an impression as a link to your site being shown to a user inside a generative AI feature on Google Search. That phrasing is important. The report is measuring exposure. It is not measuring the full answer journey.
The available dimensions are practical:
- Pages: the final canonical URL linked by the AI feature after redirects.
- Countries: the country where the search originated.
- Devices: desktop, tablet, or mobile, available for Search results.
- Dates: hourly, daily, weekly, and monthly granularity.
The Pages dimension is where I would start. It tells you which canonical URLs are being selected into supported AI features. That lets you map AI visibility by topic cluster, template type, market, and funnel stage instead of staring at a single total impression line.
Why The Totals Can Look Wrong
One reporting trap is built into the aggregation rules.
Google says the chart is aggregated by property. If two results from the same site appear in one generative AI search result, the chart counts that as one impression. If you add a URL filter, the chart switches to URL-level aggregation.
The table works differently depending on the dimension. Page data is aggregated by page. Country, device, and date data are aggregated by property. That means chart totals and table totals can differ while both are technically correct.
There are export details too. Fresh data can be preliminary. Values shown as ~ or - in the UI become zeros in downloaded data. If you feed this into Looker Studio, a sheet, or a warehouse, label those transformed values before someone builds a false trend line.
What The Report Still Does Not Answer
The new report gives SEOs a cleaner visibility layer. It does not close the AI measurement gap.
Based on Google's launch post and help docs, the report does not currently expose:
- the query or prompt that triggered the AI feature;
- the exact generated answer text;
- citation position or visual prominence;
- which competitors were cited next to you;
- clicks from AI feature impressions in the dedicated AI report;
- conversion value from those appearances;
- why Google selected one URL and ignored another.
This is the same problem I covered in the GEO attribution crisis. AI surfaces can mention a brand, show a link, answer a question, and still send very little referral traffic that analytics can classify cleanly. Search Console now confirms the exposure side for Google's supported AI features. It does not close the loop from exposure to business outcome.
How I Would Use It In An SEO Program
I would treat the report as a weekly AI visibility audit, not a daily panic dashboard.
- Confirm access and eligibility. If the report is missing, check whether the property is in the rollout and whether the site is eligible for Search generative AI features.
- Start with Pages. Export the URLs receiving AI feature impressions and map them to topic clusters, templates, markets, and funnel stage.
- Compare against regular Search performance. Look for pages with AI impressions but weak clicks, and pages with search traffic but no AI visibility.
- Split by country and device. AI visibility can land unevenly across markets and mobile versus desktop. Do not treat one global line as the whole story.
- Review the selected URLs manually. Check crawlability, snippet eligibility, canonical targets, answer clarity, first-hand evidence, structured sections, and media.
- Sample real answers. Search Console tells you that Google showed a URL. It does not tell you what the AI answer said. Review priority queries in AI Overviews and AI Mode by hand.
- Annotate changes. When you refresh a page, change internal links, add evidence, or fix technical issues, mark the date and watch the AI impression trend over the next few weeks.
The best use case is gap mapping. If a guide appears often in AI Mode but the commercial comparison page does not, that is a clue. If a support article gets AI Overview impressions while the money page is invisible, that is another clue. The report gives you the starting point for investigation.
Where This Fits With Google's AI SEO Guidance
This launch connects directly to Google's earlier documentation on optimizing for AI Overviews and AI Mode. Google has been clear that AI search eligibility sits on top of normal Search systems: crawling, indexing, quality, usefulness, media, and snippet controls.
The new report makes that guidance easier to test. Before this, a team could improve a page and then argue about whether AI visibility changed. Now, at least for properties in the rollout, the team can watch official AI feature impressions for affected URLs.
It also sharpens how we should read third-party AI visibility tools. The Ahrefs AI search benchmark showed that classic rankings, brand mentions, links, and YouTube mentions all belong in the AI visibility conversation. Prompt trackers still help outside Google, especially for ChatGPT, Perplexity, Claude, Gemini, and manual answer sampling. Inside Google Search, this new Search Console report becomes the canonical source for AI feature impressions.
That does not make prompt trackers useless. It makes their job more honest. Use Search Console for Google's official AI impression count. Use prompt sampling to inspect answer text, competitor presence, and source quality. Use analytics and CRM data to connect the work to outcomes.
What I Want Google To Add Next
Impressions are the minimum useful starting point. The version of this report that would change boardroom reporting would add:
- click data separated for AI feature impressions;
- query data, even if Google has to group it to protect privacy;
- citation position or visual prominence;
- a marker for brand mentions inside the generated answer;
- side-by-side Search and Discover AI reporting;
- clearer separation between standard result impressions and AI feature impressions in exports.
Google says it is asking for feedback through the Search Console Submit feedback link, a dedicated feedback form, and the Search Central Community. If your property has access and you need clicks or query data, that feedback channel matters.
What To Do Before Your Property Gets Access
If the report is not available in your account yet, prepare the measurement layer now.
- Tag priority URLs by topic cluster, template, market, and funnel stage.
- Keep clean exports from regular Search performance and Discover performance.
- Review snippet controls, canonicals, blocked resources, and indexability on pages you expect AI features to use.
- Build a small manual prompt-sampling workflow for your top commercial and informational topics.
- Create an annotation habit for AI-related content and technical changes.
The teams that get value from this report will be the teams that already know what each URL is supposed to do. A random table of pages with AI impressions is interesting. A mapped table by market, intent, content type, and revenue path is actionable.
The Practical Takeaway
Google's Generative AI performance reports are good news for SEOs because they turn part of AI search visibility into official Search Console data. They are also a reminder to stay precise. We can now measure official AI feature impressions for supported Google surfaces. We still cannot measure the full path from prompt to generated answer to click to revenue inside one Google report.
Use the report as the visibility layer. Use regular Search Console, analytics, logs, AI answer sampling, and revenue data as the rest of the evidence stack. The measurement gap is smaller than it was before this launch. It is not closed.
About the Author
Francisco Leon de Vivero is VP of Growth at Growing Search, with 15+ years of SEO experience across ecommerce, adult, international, and enterprise search. He worked as SEO Lead at Shopify for more than seven years and spent more than four years working on SEO for Pornhub.