Search Console's Generative AI Performance Report: What It Shows and How to Use It

Key points before the detail:

• The report counts impressions in Google's generative AI features, such as AI Overviews and AI Mode. It has no click data and no query data.
• You can group the impressions by page, country, date and, for Search, device.
• The report appears only if your site has had enough impressions in those features, and it may not appear if you have excluded your site from them.
• Chart totals and table totals can differ, because they are aggregated differently.
• The practical use is a monthly review that sets each landing page's AI impressions beside its clicks in the main Performance report. The routine is below.

What is the Generative AI performance report?

It is a Search Console report that shows how your site appears in Google's generative AI features, with one version for Search and another for Discover. Google's Search Central Blog announced it in June 2026 as dedicated views of impressions in features such as AI Overviews and AI Mode.

The same impressions were already counted in the overall performance report. Google's launch post says that data stays there, and the new report is a separate view of the generative AI part.

Search Engine Roundtable describes where to find it: an expandable tab under the main performance report. Google also says it expects to update the list of covered capabilities over time, so the Help page, not this article, is the current list.

What data does the report show?

The report shows five things: impressions, pages, countries, devices and dates. The table follows the list in Google's launch post, with the Search and Discover split taken from the two Help pages.

Google's Help page for the Search report names three jobs for it. You can see how your impressions change over time, see which pages get the most and fewest impressions, and see where they come from, for example by device or country.

What does the report leave out?

It leaves out clicks and queries, the two numbers most site owners want first. Search Engine Land reported in June 2026 that the report has no click data, and Search Engine Roundtable said in August 2026 that it still had neither clicks nor queries.

Google's answer, as Search Engine Land reported it, was that it is working with website owners on which insights help most and will add metrics over time. Search Engine Land's own view, which is opinion, is that click data should not be expected at all. Neither Help page read for this article describes a click metric.

Three smaller limits are documented. Search Console excludes data from Search Labs experiments, the 1,000-row limit of the normal performance reports applies, and values shown as ~ or - on screen turn into zeros in an export.

The table sets out which questions the report can answer. It was built by checking each common question against the fields the sources describe: where no source describes a field, the answer is No, and where the sources are silent on a split, it is marked Not documented.

Why can't you see the report?

You most likely have not had enough impressions in generative AI features, which is the cause Google's Help page gives first. The second documented cause is that you have excluded your site from Search generative AI features, in which case the site must be included again to be eligible for display.

Rollout is the other source of confusion, because the sources were written weeks apart. The table puts them in date order.

Read in date order, the sources do not conflict. The report started with a subset of sites, spread through the summer, and Google's Help page now says all websites worldwide have it as of August 31, 2026. John Mueller's comment, as quoted by Search Engine Roundtable, adds that like the Discover and News reports it is shown only when there is sufficient data.

Search Engine Land also reported that Google was testing a toggle to opt out of AI Overviews, AI Mode and AI Overviews in Discover, for a subset of UK site owners. Google said opted-out sites would receive no traffic or impressions from those features, and that the control would not be used as a ranking signal for results outside those features. Whether you have that control today is something only your Search Console can show.

How do you read the chart and the table?

The chart shows total impressions for your property over the period, and the table groups impressions by the dimension you choose. The two can legitimately disagree, because Google aggregates them differently.

• Pages groups by the final URL the feature linked to, after redirects. As in the normal report, most data is assigned to the canonical URL rather than a duplicate.
• Countries groups by the country where the search originated.
• Devices groups by desktop, tablet or mobile.
• Dates groups by day, week or month, in Pacific Time.

A search type filter sits at the top. Web: text-based covers the standard search bar, and Web: multimodal covers image searches such as Google Lens, Circle to Search, uploaded images and Chrome image search. The report draws on the Web search type of the main Performance report.

The aggregation rule is the part that trips people up. The chart counts by property, so two results from your site in one generative AI answer count as a single impression, while a table grouped by page credits each page.

Here is an illustration of that documented rule, with no real data. If one AI answer links to two of your pages, the chart adds one impression and the Pages table shows one impression on each page, so the page rows sum to two while the chart total says one. Google says the same troubleshooting tips as the normal performance report apply.

Two more details affect month-end reading. The newest data can be preliminary, shown as a dotted line, and may change within hours, and a URL filter changes the chart to aggregate by URL.

How does the Discover version differ from the Search version?

The Discover version counts a different event and offers fewer dimensions. The comparison below was built from the two Help pages, using the criteria a reader needs to avoid mixing the numbers: what counts as an impression, which dimensions exist, what the Pages dimension holds and how the chart is aggregated.

Google's Discover page adds that chart totals are accurate for the selected span and are not truncated. Daily values may be left off the chart if they do not reach a threshold, but they are still in the totals.

Do not add the two reports together. They count different events on different surfaces, and the Pages dimension means a different thing in each.

Can you find AI Overviews traffic and AI Mode performance in it?

You can find impressions from generative AI features, including AI Overviews and AI Mode, but not the traffic they send. Google's launch post names both as examples of the features covered, and the report has no clicks.

That matters for anyone searching for AI Overviews traffic in Search Console. Google says the generative AI data is part of the overall performance report, but nothing in the sources read for this article separates the clicks that came from an AI Overview from the rest. Any AI Overviews click figure taken from Search Console would be an inference.

AI Mode performance has the same limit, plus one more. The Help page lists no feature dimension, so the report gives one combined view of the covered features rather than an AI Mode view.

The report also says nothing about other AI products. It covers Google Search and Discover, so referrals from tools such as ChatGPT or Perplexity need a different source, such as your site analytics.

For the size of the click effect, the evidence sits in third-party studies, not in this report. Kanso's weekly growth checklist names two of them with their method and date.

How do you run a monthly review by landing page?

Run it once a month, on the last full calendar month, and tie every AI impression figure to a page with a job. This routine is Kanso's suggestion, built from the documented fields above, and it is not a Google procedure.

1. Confirm what you are looking at. Open the Generative AI report, note the property, the search type filter and the date range. Remember that dates run in Pacific Time.
2. Set two periods. Choose the last full month and the month before it, and compare complete months only, because the newest data can be preliminary.
3. Export the Pages table for each period. The 1,000-row limit applies, so on a large site run it by section with a URL filter.
4. Label each URL with its job. Landing page, product page, article or other. This is what turns a list of URLs into a decision.
5. Read the same URLs in the main Performance report. Record clicks and impressions for the Web search type over the same dates, and treat the comparison as directional, since the main figures include more than AI features.
6. Check countries and devices. Ask only whether the impressions come from markets and devices you serve.
7. Choose at most one change. Use the table in the next section, make the change, and note the date.
8. Log the month. Next month's comparison is only useful if this month's numbers were saved.

The only calculation is the change between months. The numbers below are made up to show the arithmetic and are not data from any property.

change = this month minus last month, so 500 minus 400 = 100; percentage change = 100 ÷ 400 × 100 = 25%

If last month is zero, the percentage is undefined, so record the raw change. The weekly version of this check lives in Kanso's weekly growth checklist.

What should you change on a page after reading the report?

Change a page only when the report points to a specific gap, and start with what Google's own guidance says decides eligibility. Google's generative AI guide says a page must be indexed and eligible to show in Search with a snippet, and the site must be included in Search generative AI features in Search Console.

Beyond eligibility, the same guide says unique, non-commodity content is likely to influence presence in generative AI search more than its other suggestions. It also says writing separate pages for every query variation, including fan-out queries, to manipulate rankings or AI responses violates its scaled content abuse policy. That is official guidance.

It lists things you can ignore for Google Search as well: llms.txt files and other special markup, chunking content into small pieces, rewriting content just for AI and seeking inauthentic mentions. The table pairs common patterns in the report with a first check. The middle column is Kanso's inference, and the third column draws on Google's documentation.

What is documented and what is analysis?

Everything in the first list comes from a named source read on October 16, 2026, and everything in the second is Kanso's reading of it.

Documented, as of October 16, 2026:

• The report shows impressions, pages, countries, devices and dates, with devices for Search only (Google's launch post).
• It had no click or query data when Search Engine Land and Search Engine Roundtable reported on it, and Google said it would add metrics over time.
• Google's Help page says the insights were rolled out to all websites worldwide as of August 31, 2026, and that the report may not show when a site lacks enough impressions or has excluded itself.
• The chart is aggregated by property and the Pages table by page, and Search and Discover count impressions differently (the two Help pages).
• Dates are in Pacific Time, exports turn ~ and - into zeros, and the 1,000-row limit applies (the Help pages).

Analysis:

• AI Overviews clicks cannot be isolated from the sources described, so any figure for them is an inference.
• A month-over-month change in AI impressions may reflect a change in which features Google counts, not only a change in your pages.
• Comparing AI impressions with main-report clicks by URL is a useful direction, not a measurement of click-through from AI answers.
• The monthly routine and the diagnosis table are Kanso's method. No one has tested them on a property for this page.

Where does Kanso's GEO Agent fit?

The GEO Agent answers a different question from this report. Search Console tells you what Google reported about your impressions, while Kanso's GEO Agent checks how readable your site is to AI systems, estimates how citable it looks and turns the gaps into actions.

Its citability view is a model estimate. Kanso does not query ChatGPT, Gemini, Claude or Perplexity, so it is not a measurement of AI citations, and it does not replace the Search Console report or give click data.

Search Console Generative AI report: frequently asked questions

Sources

• Introducing Search Generative AI performance reports in Search Console, Google Search Central Blog, June 2026
• Generative AI performance report (Search), Search Console Help
• Generative AI performance report (Discover), Search Console Help
• Optimizing your website for generative AI features on Google Search, Google Search Central
• Google Search Console AI performance reports and controls to block your content in AI responses, Search Engine Land, June 3, 2026
• Google Search Console Generative AI Performance Report Live For All, Search Engine Roundtable, August 2026