Understanding Google's AI Search Impressions: What It Means for Your Brand

What did Google just change in Search Console?

Google now reports AI visibility on its own terms instead of folding it into regular search data. On June 3, 2026, it launched a dedicated Generative AI performance report inside Search Console, pulling impressions from AI Overviews and AI Mode out of the standard Performance report and giving them a home of their own, according to Search Engine Journal.

Publishers and SEO teams had pushed for this kind of visibility since AI Overviews first started citing sources inside regular results, arguing that a page's fate inside a generative answer was otherwise invisible. Search Console's answer, after a limited beta, is a report built specifically for that question.

This is not the first time Search Console has broken out a new surface on its own terms. Discover impressions got the same treatment years earlier, arriving as a separate report rather than a line item inside standard search performance.

An impression, by Google's definition, counts an instance in which a link to a website is shown to a user inside a generative AI feature. The new report breaks that number down by page, country, device and date, and lists which URLs actually appeared inside an AI answer. What it does not include is just as telling: no query data, no click-through rate, no average position, and no detail on where inside the answer a citation landed.

The rollout started with a subset of sites while Google tested the report and gathered feedback. By August 31, 2026, the rollout had reached every website worldwide, alongside a new toggle that lets a site block its content from AI Overviews, AI Mode and AI Overviews in Discover, as reported by Search Engine Land. Opting out also costs a site any impressions or traffic that content would have sent, and Google says the choice will not affect core ranking.

Google has not said why it drew the line at impressions and stopped short of clicks or queries. The likely reason is exposure: publishing exact click-through rates on AI Overviews would show, in stark relief, how often a citation earns a visit versus how often the AI answer satisfies the user without one. An impression count avoids that comparison entirely.

Who does this change affect?

Every site with content that could plausibly answer a question is affected, whether its owner has opened the new report yet or not. B2B SaaS teams now have a native way to check whether Google's AI Overviews cite their product pages when someone searches for 'best [category] tool', or whether a competitor gets the nod instead. Local and service businesses get a first look at whether their city and category pages show up when someone asks Google's AI for the best plumber, dentist or accountant nearby.

For SaaS teams, the proof unit that matters is comparative: does Search Console show impressions on the pages built to answer 'best [category] tool' and comparison queries, or do those pages sit empty while a rival's URL fills the AI answer. For local and service businesses, the proof unit is geographic: whether a '[service] in [city]' page ever registers an AI impression at all.

The catch is that the metric is coarse. Property-level totals and page-level sums do not always match, because Google counts impressions differently depending on the level of aggregation being viewed. A link can appear as a prominent citation, as a hidden citation a user has to expand to see, or as the answer to a follow-up question nobody credits on the page itself. Comparing an AI impression to a traditional search impression, for now, means comparing two different things wearing the same label.

That ambiguity matters most for anyone reporting these numbers upward. A spike in AI impressions looks identical in a spreadsheet whether it came from ten prominent citations or a thousand buried ones, so the raw count on its own tells a manager very little about actual exposure.

How should you respond to the new data?

Treat the report as a diagnostic, not a dashboard to check every morning. A handful of moves make it useful.

1. Export the AI-visible URL list and lay it next to normal organic performance data so overlaps and gaps are easy to spot.
2. Investigate any page with AI impressions wildly out of proportion to its organic traffic. That mismatch usually points to a format Google's generative systems favor: clear structure, a direct answer near the top, a sourced stat, or a comparison table.
3. Study what the best-performing AI pages have in common, from heading structure to answer placement to whether they carry a dated statistic.
4. Keep AI impressions out of executive reporting that blends them with organic clicks. The two numbers measure different things, and blending them will overstate or understate performance depending on the week.
5. Track the trend over weeks and months, not days. Google's own guidance warns that daily AI impression counts move around enough to be misleading on their own.

What is Google still not telling you?

Plenty. The report has no query data, so there is no way to see which question triggered an impression. It has no click-through rate or average position, so there is no way to tell whether a citation actually pulled in a visitor. And it carries no detail on where in the AI answer a link sat: a prominent citation near the top reads very differently from a hidden one a user has to expand to find.

Consider two pages with identical AI impression counts. One earns a bold citation link at the top of an AI Overview, the other gets folded into a sources accordion a user must tap to expand. Both count the same in Search Console today, so a team optimizing purely on the raw number can chase the wrong outcome for months without knowing it.

The opt-out toggle raises its own question. A site that blocks AI Overviews keeps its content out of Google's generative answers, but every other engine, ChatGPT, Perplexity, Gemini and Copilot, still crawls and cites that content on its own terms. Opting out of one engine's AI does not opt a brand out of being discussed, quoted or misquoted by the rest.

Google also has not extended anything like this to the other engines answering the same questions. Each of them decides independently whether to cite a page, and none hands a site owner a native report to check. A brand chasing visibility across all of them still needs to track citation activity outside any single platform's dashboard.

Why does this matter beyond Google Search?

The bigger shift here is not the report. It is the admission behind it. Google is telling site owners, in effect, that being cited inside an AI answer is a distinct outcome from ranking on a results page, one worth measuring on its own. That lines up with the idea behind Citation Share, the percentage of relevant AI answers that cite a given brand: there is no page two in an AI answer, and a citation either happens or it does not.

The pattern shows up in practice. LeadHaste, the group Omnicite is built inside, went from zero to 1 million impressions and more than 200 AI citations in four months, while its Domain Rating moved from 1 to 24 over the same stretch (own data). None of that came from gaming any model's algorithm. It came from publishing at a pace and quality bar that let the underlying content earn a citation instead of chasing one.

A done-for-you approach to this problem looks less like waiting on each platform's dashboard and more like publishing consistently enough, and cleanly enough, that ChatGPT, Perplexity, Gemini, Copilot and AI Overviews all have a reason to cite the same brand. Search Console's new report is a start. It is not the whole picture.

Google's new report is one engine finally handing over a partial receipt. The other engines are not on the same timeline, so a brand that waits for every platform to build a dashboard will be measuring blind for a long while yet.

Sources

Source: Search Engine Journal

Google launched a dedicated Generative AI performance report in Search Console on June 3, 2026, showing AI Overviews and AI Mode impressions by page, country, device and date, without click, query or position data. Search Engine Journal, 2026-06-03

Source: Search Engine Land

Google completed the worldwide rollout of the Generative AI performance report and a new AI opt-out toggle by August 31, 2026. Search Engine Land, 2026-08-31