How to Read Search Console's Generative AI Performance Report for SaaS
Google now reports AI Overview and AI Mode impressions in Search Console. Here is what the numbers mean, what they leave out and how to use them each week.
- Desk
- Measurement
- Format
- How-to
- Reading time
- 9 min
- Published
- Nov 16, 2026
Google's generative AI performance report in Search Console shows how many impressions your pages earn inside AI Overviews and AI Mode. It lists pages, countries, devices and dates, but not clicks. Use it to see which SaaS pages Google surfaces in AI answers, then pair it with Google Analytics sessions to judge business impact.
Search Console now has a dedicated view of generative AI visibility, and most B2B software teams have not built a routine around it. This guide covers what the report contains, how Google counts its numbers, where its limits sit and a weekly review process. It is written for SaaS marketers and founders who run their own SEO and need to know whether AI answers are surfacing their pages.
What Is the Generative AI Performance Report in Search Console?
The generative AI performance report is a Search Console view that counts how often links to your site appeared inside Google's generative AI features. Google announced the report on June 3, 2026, first for a subset of sites. Per Google, the insights reached all websites worldwide as of August 31, 2026.
cekpr covers SEO and AI search for B2B software, and this report gives that second half of the job a dedicated Google data source. Many teams had relied on third-party trackers to estimate AI Overview exposure. Now the numbers come from Google itself.
The report covers four things, according to Google's announcement. The table below lists each one and what a SaaS team can do with it.
| Report element | What it shows | How a SaaS team uses it |
|---|---|---|
| Impressions | How often your URLs appeared in generative AI features | Track the trend line week over week |
| Pages | Which URLs appeared | Find the pages AI answers prefer |
| Countries | Where the searches came from | Check fit with your target markets |
| Devices | Desktop, tablet or mobile | Spot a mobile-heavy audience |
| Dates | Hourly, daily, weekly or monthly views | Line up changes with publishing dates |
Google's help page says the report includes AI Overviews and AI Mode for Search. The same page states that experiments in Search Labs are not included. A separate report covers generative AI features in Discover.
How Do You Open the Report and Fix a Missing One?
Open Search Console, choose your property and select the generative AI performance report from the Performance section. If it does not appear, the usual cause is too few impressions. Google's help page says a site that has not earned enough impressions in these features will not see the report yet.
The second cause is an exclusion setting. Google's Search generative AI control lives under Settings in Search Console. It has three choices: include your site, exclude it, or inherit the setting from a parent property.
The default is to include your site. Google's control documentation says that choosing exclusion means you receive no traffic or impressions from these features. Check this setting first on any property that shows nothing.
Domain properties and subfolder properties add a wrinkle. A child property can inherit the setting from its parent. Confirm which setting each property uses before you compare reports across a docs subdomain, a marketing site and an app subdomain.
How Does Google Count Impressions in the Report?
Google counts an impression when a link to your site is shown to a user in a generative AI feature. The chart aggregates by property, so two links from your site in one AI answer count as one impression. The page table aggregates by page, so each unique URL is counted separately.
This is the same logic Google uses in the standard performance report. Google's data-counting documentation shows an example where three pages from one site appear in one result. The property view reports one impression. The page view reports one impression for each of the three URLs.
That gap explains why chart totals and table totals rarely match. A SaaS site with a pricing page, an integrations page and a comparison page can show strong page-level numbers and a much smaller chart number. Neither is wrong.
The same documentation covers preliminary data. The newest data can still change over the next few hours, and the chart marks it with a dotted line. Do not report a number from the last day as final.
What Does the Report Not Tell a SaaS Team?
The report does not show clicks, and its documentation does not list queries. It tells you that an AI feature showed a link to your page. It does not tell you that anyone visited, or which prompt triggered the appearance. Impressions are the whole story here.
Google also notes that the usual performance report limits apply, including the 1,000-row cap on tables. A SaaS site with thousands of blog and docs URLs will not see every page. Filter by folder, such as /blog/ or /integrations/, to pull a smaller set that fits under the cap.
Two other limits matter for planning:
- Search Labs experiments are excluded, so early features will not show up.
- Impressions are not citations. A link shown in an AI answer is a different event from a brand mention in running text.
Treat the report as a visibility meter. It is not an attribution tool. For citation-level detail, Bing offers a related view, which we cover in our guide to Bing's AI Performance report and citation share.
How Do You Connect AI Impressions to Pipeline?
Match impressions from this report with organic sessions and conversions from Google Analytics, then compare the trends by page group. Google's own guidance on pairing the two tools says the metrics will not match exactly, so look at direction rather than exact totals.
Google's guide to Search Console and Google Analytics describes Search Console as the view of activity before a visit and Analytics as the view of what happens after. It also suggests using Looker Studio to chart both together. That setup fits this report well.
Use this four-step process each week:
- Export the report's page table, filtered to one folder at a time.
- Export organic sessions for the same pages and dates from Google Analytics.
- Place both in one sheet with the publish or update date of each page.
- Flag pages where impressions rose but sessions stayed flat.
The last step is the useful one. A page that is shown often in AI answers but rarely visited may answer the question so completely that nobody needs to click. Another reason is weak positioning: the AI feature may be citing the page for a minor point. Either way, the fix is editorial, so rework the opening paragraph and add a reason to click, such as a template, calculator or dataset.
Data and quotes help here too. Our write-up on adding statistics and expert quotes to SaaS articles covers study findings that point the same way.
What Does a Good Weekly Review Look Like?
A good weekly review takes under 30 minutes and answers three questions. Did impressions rise or fall? Which page group moved most? Did a content change line up with the move? Keep a dated log of changes so every spike has a likely cause attached.
Group pages into five buckets that mirror how a software buyer searches: product, pricing, integrations, comparisons and educational posts. Review each bucket as a unit. One URL swinging up or down tells you little. A whole bucket moving tells you something real.
Add a column for pages you changed that week. Google's report lets you view data by day, so you can compare the seven days before and after an edit. Google's documentation does not state how long an AI impression takes to reflect a page update. At cekpr, we advise waiting at least two full weeks before you call an edit a win.
The log also protects you from false conclusions. Impressions can fall because users search less for a topic, because AI answers cite a competitor, or because Google changed how often it shows these features. Only a log of your own changes lets you rule out the possibility that your own edit caused the move.
A related question is whether to add Google's Preferred Sources button to your pages. We weigh that choice in our analysis of Google's Preferred Sources button for B2B SaaS.
What Should You Ignore When Optimizing for Google's AI Features?
Ignore tactics that promise AI-specific shortcuts. Google's guide to optimizing for generative AI features says optimizing for these features is still SEO. It tells site owners they can skip chunking content, creating llms.txt files and chasing inauthentic mentions.
The same guide recommends non-commodity content with a unique point of view. It names the Search Console report as the way to measure results. Be wary of third-party tools that claim to use internal Google metrics.
Accuracy matters as much as visibility. A page that appears in AI answers with wrong figures hurts trust faster than a page that never appears. If your team drafts with AI help, follow the process in our guide to fact-checking AI-assisted content under Google's 2026 guidance before publishing.
Conclusion: Make the Report Part of Your Routine
The generative AI performance report gives SaaS teams a first-party answer to a question they used to guess at: is Google showing our pages in AI answers? It reports impressions only, so it works best next to Google Analytics data and a dated change log.
Start this week. Confirm the Search generative AI control is set to include your site. Export the page table for your top folder. Log your next five content edits with dates. In a month, you will have a baseline that tells you whether your work shows up in AI answers.
What does the generative AI performance report in Search Console show?
Does the generative AI performance report show clicks?
Why can't I see the generative AI performance report in my property?
How often should a SaaS team review AI impressions in Search Console?
Up next · Technical SEO How to Handle 404 Errors on a SaaS Site After Google's October 2026 Guidance join the club.
The week's best SEO and AI search posts, plus one thing to check in your own Search Console, every Thursday morning. Unsubscribe whenever.