Search Console’s Generative AI Performance Report: What It Shows, What It Hides and How to Use It

Last updated: 4 November 2026

Why this report matters

Since AI Overviews launched, site owners have had a frustrating gap. Clicks from AI features were mixed into normal web search data in the Performance report. You could see that impressions were rising and clicks falling, a pattern often discussed under zero-click search, but you could not see which part came from AI features. Third-party tools filled the gap with estimates and tracked prompts, but none of them had Google’s own numbers.

The Generative AI features performance report is the first time Google shows site owners, in its own data, how often their pages appear in its AI features. It is limited, and those limits matter, but it changes how you can talk about AI visibility. Instead of “we think Google’s AI is using our pages”, you can say “Search Console shows our pricing page appeared in AI features this many times last month”.

What the report includes

Google’s Gary Illyes has suggested that clicks and queries are unlikely to appear in this report. It is safer to plan your reporting around impressions-only data than to wait for more.

What an “impression” means here

In the normal Performance report, an impression usually means your link appeared in search results the user could see. In the AI features report, an impression means your page was shown as part of an AI feature: typically as a cited source or link within an AI Overview, an AI Mode answer or a Discover AI feature. The exact counting rules follow Google’s documentation for AI features, which can be updated, so check the help page when you start using the report.

What matters for interpretation is this: an impression tells you Google’s AI used and displayed your page. It does not tell you how prominent the link was, whether the user saw it or whether your brand was described positively. Treat it as evidence of selection, not of performance.

How to find and filter the report

1. Open Google Search Console and choose the property you want. A domain property gives the most complete view.
2. Look in the Performance section for the Generative AI features report. As a beta, it may not be visible for every property yet.
3. Set the date range. Start with the last three months to see a trend, then narrow it.
4. Switch between Search and Discover to see AI Overviews and AI Mode impressions separately from Discover AI features.
5. Add a page filter to see which URLs Google’s AI shows most often.
6. Add country and device filters if your audience is spread across markets.
7. Export the data regularly. Search Console keeps a limited history, so save monthly snapshots for your reports.

Five practical ways to use it

1. Find the pages Google’s AI already trusts

Sort pages by AI feature impressions. The pages at the top are the ones Google’s AI selects most. Look at what they have in common: clear answers near the top, specific numbers, tables, FAQ sections or fresh dates. These patterns show you what works for your site, not in theory. Our guide to refreshing content for AEO explains how to apply them to weaker pages.

2. Spot important pages that are missing

Make a list of the pages that matter most to revenue: service pages, pricing, product categories, comparison pages. Check whether they have any AI feature impressions. A key page with search impressions but no AI feature impressions is a clear content opportunity.

3. Measure the effect of changes

When you rewrite a page, add schema or update facts, note the date. Compare AI feature impressions for that page in the four weeks before and after. Because there are no clicks or queries, this before-and-after view by page is one of the most reliable ways to use the report.

4. Compare AI visibility with normal search visibility

Export normal web search impressions for the same pages and period. Pages with high search impressions but low AI feature impressions may rank well but lack the structure AI features prefer. Pages with relatively high AI feature impressions may be good citation sources even if they rank lower in classic results.

5. Report AI visibility honestly to stakeholders

Leaders often ask “are we showing up in Google’s AI?” This report lets you answer with Google’s own numbers. Pair it with the limits explained below so nobody reads impressions as traffic.

What the report cannot tell you

• Which queries triggered AI features. You will need to infer topics from the pages that appear and from your own prompt tests.
• Whether people clicked. Use GA4 referral data and landing page trends instead. See how to track AI traffic in GA4.
• AI Overviews versus AI Mode. They are one combined number. If the difference matters, use third-party tracking.
• Competitors. Search Console only shows your own property.
• Other engines. ChatGPT, Perplexity, Copilot and Claude are not included. Bing Webmaster Tools and your own tracking are needed for those.
• Brand descriptions. Impressions do not show what the AI said about you. Check that manually, as in fixing wrong AI answers about your brand.

The UK context and the opt-out toggle

The first rollout went to a subset of UK site owners. This is connected to requirements from the UK Competition and Markets Authority, which has been examining how Google’s search services treat publishers. Google is also testing a toggle that lets site owners opt out of AI features. That is a separate control from training preferences, which we cover in our guide to the new AI training opt-out in robots.txt.

Opting out of AI features is a serious decision. It would remove your pages from a growing part of Google’s results. For most businesses that want to be found, the better path is to improve how pages appear in AI features rather than leave them. Publishers whose business model depends on clicks may see it differently, and the new report gives them evidence to make that decision.

Building an AI visibility dashboard with this data

The report becomes much more useful when combined with other sources. Here is the structure we use in client reporting.

Our AEO reporting template shows how to lay this out month by month, and our list of AEO KPIs explains which metrics matter at each stage.

A monthly routine for the report

1. Export last month’s data for Search and Discover, by page, and save it.
2. List the top 20 pages by AI feature impressions and note any new entries or drops.
3. Check your priority pages. Mark which ones have AI feature impressions and which do not.
4. Match changes to dates. Compare with your content change log and any Google updates.
5. Check a sample of answers manually. Search a few topics in AI Mode to see how your pages are used.
6. Add GA4 numbers for AI referral sessions and conversions.
7. Write three lines of insight: what went up, what went down and what you will change next month.

How to improve AI feature impressions

The report shows outcomes. These are the inputs that usually move them, based on Google’s guidance for AI features and on what we see working.

• Be eligible first. Google says there are no special requirements for AI features beyond being indexed and eligible to show a snippet. Check that key pages are indexed, not blocked by nosnippet or noindex, and readable without JavaScript.
• Answer the question early. A clear one or two sentence answer near the top makes a page easy to use as a source.
• Use structure. Headings that match questions, short paragraphs, tables and lists help engines pull the right part.
• Show specifics. Prices, ranges, dates, steps and named examples are more useful to an AI answer than general claims.
• Keep facts fresh. Update pages when prices, features or rules change, and show a visible updated date.
• Strengthen the entity. Organization schema, consistent business details and mentions on trusted sites help Google understand who you are.

For a full checklist, see AI Overviews optimisation and the technical AEO checklist.

Example: reading the report for a service business

This is an illustration, not a client result. Imagine an accounting firm in Ahmedabad looks at three months of data.

• The “GST registration process” guide has the most AI feature impressions, rising each month.
• The “GST registration fees” page has strong normal search impressions but almost no AI feature impressions.
• The homepage and “About” page have none, which is normal for most businesses.
• Discover AI feature impressions are close to zero, which is expected for a service firm.

What should the firm do? The process guide is working: it answers a clear question with steps. The fees page ranks but is not used by AI, so the firm checks it and finds the fees are in an image, with no text answer or table. It adds a short answer at the top (“Government fees for GST registration are…”), a table of costs by business type, an FAQ and an updated date. Four weeks later, it compares AI feature impressions for the page. Whatever the result, the firm now has evidence instead of guesses.

How it compares with other data sources

No single source is enough. The Search Console report is the most trustworthy for Google’s AI features, and the weakest for everything else.

Common misreadings

• “Impressions fell, so AI traffic fell.” Not necessarily. Impressions and clicks can move differently. Check GA4.
• “We have no AI impressions, so Google’s AI ignores us.” Check whether the report has rolled out to your property, and check the right property type and date range.
• “Our competitor must have more.” You cannot see competitor data here. Use prompt tracking for comparisons.
• “AI Mode is doing better than AI Overviews.” The report cannot tell you that, because they are combined.

What to ask your SEO or AEO provider

1. Have you checked whether the Generative AI features report is available for our property?
2. Which of our priority pages appear in AI features, and which do not?
3. How are you separating AI feature impressions from clicks in your reports?
4. Which changes did you make last month, and what happened to AI feature impressions for those pages?
5. How are you measuring visibility in ChatGPT, Perplexity and Copilot, which this report does not cover?

Where this is heading

Search Console reports usually start limited and gain features. This one may gain more filters or clearer definitions as it leaves beta. But the absence of clicks and queries looks deliberate, so plan for an impressions-only world. Combine Google’s data with GA4, Bing Webmaster Tools and cross-engine prompt tracking, and you will have a far better view of AI visibility than most businesses have today. If you want a quick starting point, run our free AI visibility checker to make sure nothing technical is stopping Google’s AI from using your pages.

Setting up a page group view for AEO

Looking at pages one by one gets slow on larger sites. A better approach is to group pages by purpose and look at AI feature impressions for each group. Search Console filters by URL, so a clear URL structure makes this much easier.

Use the “URL contains” filter for each pattern, export the totals each month and track the share of AI feature impressions each group receives. If guides take almost all impressions while service pages get none, your commercial pages probably need clearer answers, prices and comparisons.

Reading trends after Google updates

Google changes its AI features often: new layouts, more or fewer links, expansions to new countries and model updates. These changes can move impressions for every site at once. Before you credit or blame your own work for a change, check three things.

1. Is the change site-wide or limited to a few pages? A site-wide shift on one date usually points to a Google change.
2. Did Google announce anything? Check the Search Central blog and reliable industry news for the same week.
3. Did your content or technical setup change? Compare with your change log, deployments and robots.txt history.

Keep a simple timeline in your report with Google announcements and your own changes side by side. Over a few months it becomes much easier to separate what you control from what you do not.

Explaining the report to non-specialists

Business owners and leadership teams do not need every detail. They need a clear, honest summary they can act on. A short explanation like this works well in a monthly report:

“Google’s AI features showed our pages about X times this month, up or down Y% from last month. The pages Google’s AI used most were A, B and C. Google does not tell us how many people clicked from these AI answers or what they searched, so we combine this with website analytics and our own AI answer tests. Next month we will improve pages D and E, which matter for sales but do not yet appear in AI features.”

Avoid turning impressions into revenue estimates. Show impressions as a visibility measure, show GA4 conversions as the business result and explain the gap between them. Leaders tend to trust reports more when the limits are stated up front.

Checklist before you rely on the numbers

• Confirm you are looking at the right property, ideally a domain property covering all subdomains and protocols.
• Check the date range and whether partial days are included.
• Separate Search and Discover views before drawing conclusions.
• Note when the report first became available for your property, so early weeks are not mistaken for a sudden rise.
• Record any Google announcements that could explain site-wide changes.
• Keep exports, because history in Search Console is limited.

Frequently asked questions

Sources

• Generative AI performance reports in Search Console, Google Search Central Blog
• GSC AI performance reports explained, CrawlRaven
• AI features and your website, Google Search Central
• Search news recap, 2 November 2026, Search Engine Roundtable