AI visibility in Search Console: using the new reports

Since 31 August 2026, Google Search Console has shown all websites worldwide how often their pages appeared in AI Overviews and AI Mode (Google Search Central). For online shops, this closes a gap that until now could only be filled with their own samples: AI visibility in Search Console can now be analyzed by page, country, device and time period. The new report only measures impressions, though. Clicks, queries and positions remain exclusively in the overall performance report. This guide explains what the reports count, how to configure them, how to compare AI impressions with the overall report and where the data ends. If you would rather not set up the analysis yourself, it is part of our SEO services.

What the new reports show

Google announced the generative AI performance reports on 3 June 2026 and initially rolled them out to a subset of websites in order to test them and gather feedback (Google Search Central). Since 31 August 2026 they have been available to all websites worldwide (Google Search Central). There is one report for Google Search and a separate one for Discover. The Search report covers impressions from AI Overviews and AI Mode; Google expects to update the list of covered features as it develops Search (Search Console Help).

What is new is not the data itself but its separation. Impressions from AI features were already included in the performance report (Google Search Central), within the "Web" search type (Search Console Help). Anyone analyzing a page there saw the sum of classic results and AI answers without being able to tell them apart. The new report is an additional view of the same data. Google puts it this way:

This data is included in the overall performance report, where it will continue to be tracked to give site owners an overview of the overall visibility of their site in Google Search.

Google Search Central Blog: Introducing Search Generative AI performance reports in Search Console

In practice this means: the new report answers the question where and how often your pages appear in AI answers from Google. Whether that turns into visits is a question it does not answer. The following comparison, compiled from the Search Console Help, shows which report answers which question.

What counts as an AI impression

An impression in the new report indicates how many times links to your site were shown to a user in a generative AI feature on Google Search (Search Console Help). Standard impression rules apply to AI Overviews and AI Mode (Search Console Help). In general, an impression is counted whenever an item appears on the current page of results, even if it was not scrolled into view; inside elements that expand or scroll independently, an item typically has to come into view first (Search Console Help). An impression therefore means shown, not read.

Position follows its own rules as well. An AI Overview occupies a single position in the search results, and all links in it are assigned that same position (Search Console Help). In AI Mode, position follows the methodology of a normal results page; if someone asks a follow-up question, that counts as a new query, and all data in the new response is attributed to that query (Search Console Help). In the overall report, a page can therefore show a good average position even though it was only one of many links in an AI Overview.

For every total, the type of aggregation matters. The chart aggregates by property: if two results from the same site appear in an AI feature, they count as a single impression in the chart total (Search Console Help). In the table it depends on the dimension: countries, devices and dates are aggregated by property, pages by page (Search Console Help). Chart and table totals can therefore differ without anything having been measured incorrectly. Just as important is what the report does not contain:

• Search Labs experiments: Search Console does not include data from these experiments because they are still in active development (Search Console Help).
• Queries: The help page lists pages, countries, dates and devices as dimensions (Search Console Help). Which query triggered an AI answer with your link is not shown.
• Clicks, click-through rate and position: The announcement lists impressions, pages, countries, devices and dates as the content of the reports (Google Search Central). The other metrics remain reserved for the overall report.
• Discover: AI features in Discover have their own report. There, all data is aggregated by page, and a link only counts once it has been scrolled into view (Search Console Help).

Opening and configuring the report

You will find the report among the performance reports in Search Console, separately for Google Search and for Discover (Search Console Help). If it is missing for your property, Google says this usually happens when the site has not yet received enough impressions in AI features (Search Console Help). Another reason can be a control that has also been available to all websites since 31 August 2026: excluding the site from Search generative AI features (Search Console Help). Before you compare figures, check the report settings. These settings change the result the most:

1. Date range and granularity: Performance can be monitored with hourly, daily, weekly and monthly granularity (Google Search Central). Weeks run from Sunday to Saturday, and months are calendar months (Search Console Help).
2. Search type: The filter separates "Web: text-based" from "Web: multimodal". The multimodal type covers searches with images, for example via Google Lens, Circle to Search on Android, uploaded images and Chrome image search (Search Console Help). For shops with image-driven ranges such as fashion, furniture or spare parts, this deserves a separate look.
3. Dimension: pages, countries, devices or dates. The choice determines how the table aggregates and which totals you can compare with each other.
4. Filter: A URL filter narrows the view to directories or page types. With a URL filter, the chart also aggregates by URL instead of by property (Search Console Help).

Analyzing pages: which shop pages appear in AI answers

For shops, the pages dimension is the most productive one. It groups data by the final URL linked by an AI feature after any redirects; as in the performance report, most data is assigned to the canonical URL, not to a duplicate URL (Search Console Help). For shops with variants this means that impressions of a color or size variant usually appear under the canonical page. How cleanly your canonicals are set therefore also determines how meaningful the page list is. Our article on product variants and duplicate content explains the background.

The table shows 1,000 rows at most, because the usual data limitations of the performance report also apply to the AI report (Search Console Help). Rare rows can be missing from the table while still being included in the chart total (Search Console Help). For large catalogs it is therefore worth filtering by page type rather than reading the whole list. In most shops, page types can be separated by path patterns, for example a dedicated directory for the guide section or a fixed pattern for product URLs. This grouping is a good starting point:

For all page types, what Google states about AI features in general applies: there are no additional requirements and no special optimizations needed to appear in AI Overviews or AI Mode. A page must be indexed and eligible to be shown with a snippet (Google Search Central). How to prepare product pages for AI answers is covered in our article on AI Overviews in shopping. If important pages are missing entirely, the cause may lie earlier, in the index, for example under “Crawled – currently not indexed”.

AI share per page: comparing with the overall report

The new report only becomes really useful next to the overall report. Both refer to the same search type: the AI report includes data from the "Web" search type in the performance report (Search Console Help). If you choose the same date range, the same search type and the pages dimension in both reports, both tables are aggregated by page and can be joined by URL. Dividing a page's AI impressions by its impressions in the overall report gives you a metric of your own: the AI share of that page's impressions. Google does not report this metric; it is a calculation from both exports. Here is how to proceed:

1. Export both reports: same date range, same search type, pages dimension. The export contains both chart and table data (Search Console Help).
2. Join by URL: Only pages that appear in both tables produce a share. If a page is missing from the AI export, it did not necessarily have zero AI impressions, because this table is limited as well.
3. Calculate the share: AI impressions divided by impressions in the overall report, per page and month.
4. Add clicks alongside: Clicks and click-through rate come exclusively from the overall report. They show whether a page is losing visits despite rising AI impressions.
5. Mark what was not measured: Values shown in the report as "~" or "-" become zeros in the export (Search Console Help). They belong in a separate "not measured" column, not in the calculation.

The comparison reveals recurring patterns:

The patterns are indications, not proof; a falling click share can also have seasonal reasons. In our article on the traffic strategy for Google AI Mode, Search Console was still an incomplete mirror of AI Mode; with its own report, that changes at least for impressions. Samples with your own prompts, as described in our article on LLM visibility and share of model, remain necessary for other assistants.

Reading countries, devices and trends over time

The countries dimension groups data by the country where the search originated (Search Console Help). For shops in German-speaking markets, this is the quickest way to check whether AI answers in Germany, Austria and Switzerland point to the same pages. If searches from Austria land on the German version even though there is a separate one, it is worth checking whether the language and country versions are marked up properly. Our article on hreflang in online shops explains how.

The devices dimension distinguishes desktop, tablet and mobile (Search Console Help). It mainly helps with prioritizing: a page whose AI impressions occur mostly on mobile devices should load quickly on a smartphone and continue the answer outlined in the AI feature without detours.

Trends over time show whether changes work. Because page changes, template switches and relaunches do not appear in the report by themselves, the Search Console annotation feature is useful: custom annotations mark events such as launching a new feature or fixing a bug directly on the chart of the performance report (Search Console Help). Each property can have up to 200 annotations, each with up to 120 characters; annotations older than 500 days are deleted automatically (Search Console Help).

The new control: including or excluding AI features

At the same time as the reports, a control directly related to them became generally available. Since 31 August 2026, all websites worldwide can decide in Search Console whether their content may appear in AI Overviews, AI Mode and generative AI features in Discover (Search Console Help). The control is located under "Settings > Search generative AI" and offers the following options:

• Include: Content can appear as links and help to ground AI answers; the site can receive impressions and traffic from these features. This is the default (Search Console Help).
• Exclude: The site's content and links do not appear in these features, and the site receives no traffic or impressions from them (Search Console Help).
• Inherit: The property follows the control of its parent property, for example a directory property following the domain property (Search Console Help).

The limits of the control are important. According to Google, it is not a ranking or inclusion signal for other parts of Search, and it does not override choices about services such as Merchant Center and Google Ads (Search Console Help). It does not affect AI training either: for that, Google refers to Google-Extended, and anyone who wants to remove content from Search entirely uses noindex (Search Console Help). How to control training crawlers and other bots beyond that is explained in our article on controlling AI crawlers.

For shops with several properties, inheritance matters: by default, a property inherits the control from its closest parent that has stopped inheriting, otherwise from the domain property (Search Console Help). A guide directory, for example, could be treated differently from the rest of the shop.

Knowing the limits of the reports

The report is a step forward, but not a complete picture. For a reliable analysis, keep these limits in mind:

• Google only: The report covers features of Google Search, the Discover report those of Discover. Answers from other assistants are not included.
• No queries: Which question led to the impression remains open. The queries for a page in the overall report only give an indication, because they mix classic results and AI answers.
• No dedicated API type: The reference for the Search Analytics API lists discover, googleNews, news, image, video and web as types; it has no separate type for generative AI features (Search Console API reference). For regular analyses, the export from the report remains the way.
• Coarse positions: Within an AI Overview, Search Console does not distinguish places; all links carry the same position (Search Console Help).

Your own data therefore remains important: web analytics shows what happens after the click, and the server's access logs show which crawlers retrieve which pages. How to analyze the latter is explained in our article on log file analysis for shops.

A monthly routine for shops

To make sure the reports are not looked at only once, a fixed routine helps. This is the routine we suggest for monthly analyses:

1. Wait for the end of the month: analyze a few days after the month ends so that no preliminary data slips into the figures.
2. Filter page types: guides, categories, products and service pages, each with its own URL filter.
3. Join the exports: AI report and overall report per page, calculate the AI share, mark what was not measured.
4. Check conspicuous pages: pages with a sharply rising AI share and falling clicks, as well as pages without AI impressions despite high overall impressions.
5. Annotate changes: record a relaunch, template change, new content or a changed AI control as an annotation in the performance report.
6. Derive measures: add content, maintain product data, fix technical causes and measure against the same grid the following month.

For product data, what applies to Search as a whole applies here too: structured data should match the visible text (Google Search Central). How attributes and descriptions can be kept consistent in large catalogs is described under AI data enrichment. The technical basics are covered in our guide to structured data for online shops.

From impression counter to decision

The new reports use Google's own data to show which shop pages appear in AI answers on Google Search. Decisions, however, need a second look at the overall report, at your own web analytics and at the shop itself. As part of our SEO services, we set up the analysis per page type, combine the exports and derive measures, from content and structured data to technical causes in templates and extensions, which we regularly maintain as a Shopware agency.

Whether excluding individual sections is worthwhile or whether more AI impressions are the goal is a trade-off we make together with you based on your figures. Contact us for consulting on your analysis.