Search Console AI Report: How to Read Google’s Generative AI Data

Dominik Breitbach Founder and Lead SEO Strategist at taismo

Dominik Breitbach founded taismo GmbH, an SEO and GEO agency from Munich, and works as lead SEO strategist on SEO retainers and visibility in AI answers. He reads the generative AI report for client properties every month and built the regex filters described here.

⏱ Reading time: 14 min🔄 Last updated: 4 October 2026

The Search Console AI report, officially called the generative AI performance report, counts how often links to your site were shown in Google’s AI Overviews and AI Mode. Google launched it on 3 June 2026 and rolled it out to all websites worldwide on 31 August 2026. It reports exactly one metric, impressions, grouped by page, country, device and date. Clicks and the questions behind those impressions sit in the standard Performance report, and this guide shows you how to read both together.

👉 Want to know where your site stands in classic and AI search? Start with an SEO audit

The Search Console AI report counts impressions in AI Overviews and AI Mode

Google’s own name for the feature is the generative AI performance report. It lives in Google Search Console next to the familiar Performance report and isolates one slice of your search data: The moments when a link to your site appeared inside an AI Overview or an AI Mode answer.

That slice was never missing from your account. Google documents that AI features are counted in the regular Performance report, inside the “Web” search type. The new report pulls them out into a dedicated view. According to the Google Help Center, you use it for three jobs:

1. Trend: See how your organic impressions from generative AI features change over time.
2. Pages: See which URLs get the highest and the lowest impressions in AI features.
3. Origin: Understand where AI impressions come from, by country and by device.

Two surfaces feed the report today: AI Overviews and AI Mode. Experiments in Search Labs are excluded because Google still treats them as development work. Google Discover has its own generative AI performance report, so AI features in the Discover feed never show up in the Search version.

The value of the report is simple to state. Before June 2026, a citation in an AI Overview and a blue link at the same position produced the same kind of impression in your data. Now you can see which of your pages Google actually uses as supporting links in its AI answers. That is the first Google data point for answer engine optimization (AEO), the discipline also called generative engine optimization or GEO.

Where to find the generative AI performance report

The report needs no setup beyond a verified property. Four steps get you to the data:

• Step 1: Open the property in Search Console.
• Step 2: Open Performance in the left navigation and select the Generative AI entry for Search.
• Step 3: Set the date range. Our property showed the last three months when we first opened it, so the report starts with a short history.
• Step 4: Pick a dimension tab below the chart: Pages, Countries, Dates or Devices.

Google names three reasons why a property shows no report. Access was rolled out gradually until 31 August 2026. A site may simply have too few impressions in AI features to produce data. Or the site was excluded from generative AI features through the Search generative AI control, which we cover further down. If the report is empty for a site that ranks well, check that setting first.

Four dimensions, two search types and one metric

The default view shows impressions for your whole property. Every other view comes from choosing a dimension tab and a search type filter. The table below explains what each option groups and what it is good for in practice.

The usual limits of the Performance report apply here too. The table shows at most 1,000 rows, so large sites see a truncated page list. The newest data can be preliminary and appears as a dotted line in the chart. When you export, values shown as ~ or – become zeros in the file. Chart totals and table totals can differ because the chart aggregates by property and the page table aggregates by page.

How Google counts an impression in AI Overviews and AI Mode

An impression in this report means that a link to your site was shown to a user in a generative AI feature on Google Search. Google applies its standard impression rules. Five counting rules from the Google documentation change how you should read the numbers:

1. Property aggregation in the chart: Two links from the same site in one AI answer count as one impression in the chart total. Add a URL filter, and the chart aggregates by URL instead.
2. Page aggregation in the table: Every unique URL in the answer counts separately in the Pages tab. That is why the sum of the page table can exceed the chart total.
3. One position for the whole AI Overview: In the regular Performance report, an AI Overview occupies a single position, and every link inside it receives that same position. A citation on a topic where your blue link ranks on page two can therefore lift or lower your average position.
4. Follow-up questions in AI Mode are new queries: When a user asks a follow-up question in AI Mode, Google counts impressions, position and clicks for that new query.
5. Clicks count in the Performance report: A click on a link in an AI Overview or in AI Mode counts as a click in the regular report. The AI report itself shows no clicks.

Dates follow Pacific Time. For a company in Europe, the first and the last hours of a calendar month therefore belong to a different day than in your analytics tool. The effect is small on a monthly total and large on a single day, which is one more reason to read the report in weeks and months.

Clicks, queries and ChatGPT come from other sources

The report answers one question very well: Which of my pages does Google show in its AI answers, and how often? Three further questions need other data, and it helps to know the source for each before you start a report.

The third row matters most for B2B companies. Buyers often research a vendor shortlist in ChatGPT before they ever type a brand name into Google, and that research stays invisible in Search Console. Our guide on AI search visibility explains how to combine Google data, prompt panels, referral traffic and server logs into one measurement system. This article stays with the Google report and gets the most out of it.

What 42,266 AI impressions on our own site showed

We exported the report for taismo.de at the start of October 2026. It covered 30 June to 29 September 2026, 92 days. The property collected 42,266 impressions in AI Overviews and AI Mode, against 340,804 impressions in classic web search over the same period. That is roughly 12 AI impressions per 100 web impressions. 82 percent of the AI impressions came from Germany, and 81 percent from desktop devices, which fits an audience that researches service providers at work.

The total looked impressive, and it was misleading until we split it by page. Three lessons came out of that split.

Lesson 1: One page can carry half of your AI impressions

47.5 percent of all AI impressions came from a single URL, our English glossary entry on AI Mode. It started to appear for the query “ai mode” in late August, peaked around 8 September and faded afterwards. In web search the same page collected 38,168 impressions at positions 4 to 8 and earned 5 clicks. People who type “ai mode” want to use the feature, so they rarely click a definition. The position stayed stable while the impressions fell, which marks a fading wave of demand. Rankings held, and the spike belongs out of any visibility claim.

Lesson 2: The baseline tells the real story

Without that page, 22,183 AI impressions remained. In July and August the daily values ranged from 50 to 280. At the end of September, after the wave had passed, they ranged from 300 to 600. The baseline roughly doubled. That growth is the result we report, and it would have disappeared behind the spike in a property total.

Lesson 3: Comparison content earns AI impressions at a much higher rate

We divided AI impressions by web impressions for each group of pages. A comparison article that ranks agencies earned 39 AI impressions per 100 web impressions. Our glossary pages earned 5. Google evidently pulls comparison content into answers for vendor questions far more often than definitions. For your own site, this ratio is the most useful number in the report, because it shows which content types Google trusts as sources, independent of how much traffic a page gets.

Find the questions behind your AI impressions in the Performance report

The AI report has no query dimension. The standard Performance report has one, and since AI Mode arrived it contains a new kind of query: Long, conversational questions in full sentences. Google documents that AI Overviews and AI Mode may use a query fan-out technique, issuing several related searches to build one answer. Google does not label any query as AI generated, but these long questions are the closest public trace of what people and tools ask.

Search Console supports regular expressions in RE2 syntax in the query filter. Two filters surface the questions:

• Long queries: Filter queries with the custom regex ^(\S+\s+){7,}\S+$. It keeps queries with eight or more words.
• Question starts: Filter with ^(who|what|which|how|why|where|can|should|is|are|does|best|compare)\b. It keeps queries that open like a question.

On taismo.de the long query filter returned 2,396 combinations of query and page with 18,034 impressions over the same period, 5.7 percent of all web impressions, and zero clicks. About half asked for a provider, with phrasing like “which agency” or “recommend”. Some carried obvious templates from tracking tools, such as “list only real, existing brands” or “my location is germany”. Those rows are monitoring traffic from tools and agents.

Even the tool traffic is useful. It shows which questions your market watches. When competitors’ tools ask for “alternatives to” a known agency and Google answers with your page, that question deserves a place in your own prompt list. In our GEO audit we turn exactly these documented questions into the prompt set we measure, which beats any list we could invent at a desk.

Include, exclude or inherit: The Search generative AI control

The report has a sibling in Search Console. Under Settings you find the Search generative AI control, also rolled out to all sites on 31 August 2026. It offers three options:

1. Include (default): Your content can appear as links in AI Overviews, AI Mode and generative AI features in Discover, and it can help ground AI responses.
2. Exclude: Your links and content disappear from these features, including grounding. You receive no traffic and no impressions from them.
3. Inherit from parent: The property follows the setting of its parent property, which is the default for properties below a domain property.

Four details from the Google documentation matter before anyone touches the switch. Exclusion takes effect within one to two days, and some content takes longer because of caching. Google states that the control is not a ranking or inclusion signal for other parts of Search. It does not affect AI training, which is governed by Google-Extended. And only noindex removes a page from Google Search entirely.

Google itself recommends the AI report as the tool to estimate what a change would cost. Our position for most businesses is clear: Stay included. A citation in an AI answer puts your brand in front of buyers at the moment they form a shortlist, and other sites fill the slot you leave empty. Our generative engine optimization services start from that inclusion and work on being the cited source.

Four patterns in the AI report and the decision each one calls for

Numbers only matter when they change what you do next. These four patterns appear again and again when we read the report for client sites.

The first pattern is the familiar effect of zero-click search, and our article on AI Overviews and SEO covers the click side in depth. The second pattern is where the report pays for itself. A page that Google already cites is your best template, and the writing rules in our playbook on how to get cited by AI help you apply its pattern to other pages.

The weight of each pattern depends on your business model. Publishers and guide sites feel the first pattern hardest, because answering questions is their product. B2B companies and service providers gain most from the second and third, because their buyers research long before they call. Shops see the effect on buying guides, while product and price searches stay closer to the classic click.

Put AI impressions into your monthly SEO report without overselling them

Management wants one answer: Are we visible in AI answers, and is it getting better? Three lines in your report deliver that answer cleanly:

1. Classic visibility: Web impressions, clicks and average position for the month.
2. AI visibility: AI impressions for the month, the five pages that carry them and the baseline without outliers.
3. Action: One decision per pattern from the table above, with an owner.

Two calculations stay out of the report. The first is a click-through rate built from AI impressions and web clicks, because the report holds no clicks and the web clicks include classic results. The second is a property total without a page split, because a single wave can double it, as our own data shows.

Plan the data flow too. When we checked in October 2026, the AI report was not available through the Search Console API, so the monthly export runs from the interface into a spreadsheet. The query side runs through the API, which makes the conversational queries easy to automate for many properties at once.

A 30 minute monthly routine for the Search Console AI report

Five steps, the same filters every month, and the report turns from a curiosity into a steering tool:

• Step 1, export: Export the AI report with the Pages and Dates tabs into your sheet. Keep every month.
• Step 2, split: Sort pages by impressions and mark every page with a sudden jump. Calculate the baseline without them.
• Step 3, compare: Put AI impressions next to web impressions per page and per page group. Watch the ratio first.
• Step 4, read: Run the two regex filters in the Performance report and save the new provider questions for your prompt list.
• Step 5, decide: Assign one action to each pattern you found and check its effect in next month’s export.

The routine works because it repeats. One export is a snapshot; three exports show a direction; six exports show whether your content decisions changed how Google uses your site.

Frequently asked questions about the Search Console AI report

How do I see AI Overview data in Google Search Console?

Open your property, go to Performance and select the Generative AI report for Search. It shows impressions from AI Overviews and AI Mode by page, country, device and date.

Does Search Console show clicks from AI Overviews?

The generative AI report shows impressions only. Clicks on links in AI Overviews and AI Mode count in the regular Performance report under the Web search type, mixed with classic clicks.

Can I see which queries triggered my AI Overview impressions?

The AI report has no query dimension. The closest signal is the query table of the regular Performance report, filtered with a regex for long, conversational queries.

Why don’t I see the generative AI performance report?

Google names three reasons: Your property may not have had access before the worldwide rollout on 31 August 2026, your site may have too few AI impressions, or it may be excluded through the Search generative AI control.

Is the Search Console AI report available in the API?

When we checked in October 2026, it was not. You export the report from the interface; the regular Performance data, including long queries, remains available through the API.

Does excluding my site from AI features hurt my rankings?

Google states that the Search generative AI control is not a ranking or inclusion signal for other parts of Search. Your pages simply stop appearing in AI Overviews, AI Mode and generative AI features in Discover.

Sources

• Google Search Central Blog: “Introducing Search Generative AI performance reports in Search Console”, Hillel Maoz and Moshe Samet, Google, 3 June 2026, updated 31 August 2026.
• Search Console Help: “Generative AI performance report (Search)”, Google, accessed 4 October 2026.
• Search Console Help: “Search generative AI control”, Google, accessed 4 October 2026.
• Search Console Help: “What are impressions, position, and clicks?”, sections AI Overviews and AI Mode, Google, accessed 4 October 2026.
• Search Console Help: “Performance report (Search results): About the data”, Google, accessed 4 October 2026.
• Google Search Central: “AI features and your website”, Google, accessed 4 October 2026.
• Google Search Central Blog: “Search Console regex filters update and quick tips”, Google, June 2021.