Search Console's AI performance report: how to read it

Search Console's new Generative AI performance report tells you how often your pages showed up in AI Overviews and AI Mode. It does not tell you whether anyone clicked. Google shipped the report on June 3, 2026, and it's the first first-party visibility data GSC has ever offered for the AI-generated slice of search, which is exactly why it's easy to over-read. Here's what the report actually contains, what it leaves out, and how it sits next to the paid AI citation tracking stack most teams already run.

What Search Console's Generative AI performance report actually shows

The report shows impressions, and only impressions, for pages that appeared in AI Overviews or AI Mode on Google Search, broken down across five dimensions with hourly-to-monthly granularity (Google Search Central). It's a visibility count, not a traffic count. Treat it as answering "did Google's AI show my page," never "did that turn into a visit."

Where to find it

The report lives inside Search Console under Performance, next to the standard Search results view, as a new Generative AI tab (Search Engine Land). If you've used the Performance report before, the interface will feel immediately familiar: same date-range picker, same comparison toggle, same CSV export in the top corner. Google didn't build a new tool, it split an existing one. The one habit that won't carry over: if you're used to sorting the Search results report by CTR to find underperforming pages, there's no CTR column here to sort by, the dimension tabs on the left read Pages, Countries, Devices, and Dates, with impressions as the only metric behind each.

The five dimensions you can slice the data by

You can break the data down by impressions, pages, countries, devices (Search only), and dates, with hourly, daily, weekly, and monthly views (Google Search Central). That's enough to answer real questions: which pages are getting pulled into AI answers, which countries drive that visibility, and whether a spike lines up with a specific publish date or a Google update. It's not enough to answer which query triggered any of it, which is the gap below.

What's missing from the report

The report currently excludes clicks, click-through rate, and query-level data entirely (Search Console Help). This is the single fact to hold onto before you build a dashboard around this data. As one analysis put it plainly, the report "answers 'are we showing up?' not 'is it sending us any traffic?'" (Zinc Digital). Google itself has framed the current five dimensions as a starting point: the team is "continuing to work with website owners to understand what insights and data would be most helpful to inform their strategies, such as adding additional metrics over time" (PPC Land). Read that as an admission that clicks and queries are coming, just not yet.

The impression threshold that hides the report for smaller sites

Google is rolling the report out to a subset of website owners first, "allowing for thorough testing before rolling it further" (Search Console Help), and a site also needs to clear a minimum AI-impression volume before the tab becomes visible. If your Generative AI tab is empty or missing, that's the two most likely reasons, not a broken account. The rollout started with a subset of owners in the UK before expanding (Brodie Clark Consulting), so global availability is also still catching up for plenty of sites that would otherwise clear the threshold.

How to read impressions-only data without over-interpreting it

An impression in this report means Google's AI layer generated an answer that referenced your page. It does not mean a person read your name, and it does not mean anyone visited your site. Keep those three ideas separate every time you look at the chart, because the report's whole design invites you to blur them.

This isn't new data, it's a new view

The AI impressions in this report were already counted inside your Search Console totals before June 2026; Google split them into a dedicated view rather than adding new measurement (Zinc Digital). That matters for two reasons. First, your overall impression numbers on the standard Search results report won't jump because of this launch, since nothing new is being measured, it's just visible in a new place. Second, if you've been tracking total impressions for months, you can go back and estimate how much of that figure was already AI-driven, which is a more honest starting point than treating July 2026 as day one.

An impression isn't a citation and isn't a click

These three words get used interchangeably in AI-search coverage, and they shouldn't be. An impression means your page was one of the sources Google's generative system drew from to build an answer. A citation, in the stricter sense researchers use, means your brand or URL was visibly named or linked inside that answer. A click means someone acted on it. GSC's report only gives you the first of the three, and the gap between "impression" and "click" is not small. Seer Interactive's analysis of 3,119 informational queries (25.1 million organic impressions, June 2024 through September 2025) found organic CTR fell 61%, from 1.76% to 0.61%, on queries where an AI Overview appeared, while paid CTR fell 68% over the same stretch (Seer Interactive). An impression count climbing steadily can sit right on top of a click rate that's quietly collapsing, and the GSC report alone will never show you that.

The same research found brands actually cited within an AI Overview posted a 35% higher organic CTR and a 91% higher paid CTR than uncited brands on the same query set (0.70% versus 0.52% organic, 7.89% versus 4.14% paid) (Seer Interactive). That's exactly the distinction the new report can't draw for you: it will tell you an AI feature included your page, but not whether you were the named, cited source with a meaningfully better CTR or one of several pages the model briefly touched and didn't surface.

AI Overviews and AI Mode are combined in one report, not split out

Both surfaces roll into a single impressions figure, with no separate chart or table for each (Search Console Help). That's a real limitation, because the two surfaces don't behave the same way. Ahrefs analyzed 540,000 query pairs and found AI Overviews and AI Mode overlap on only 13.7% of the URLs they cite for the same query. A page can be a strong AI Overview source and nearly invisible in AI Mode, or the reverse, and a single blended impressions line in GSC has no way to tell you which. If your combined number is flat but you suspect one surface is carrying it, a manual prompt log across both is currently the only way to separate them.

What a rising or falling trend actually tells you

A rising trend tells you Google's AI systems are pulling from your pages more often, full stop. It doesn't tell you why. It could mean you published something that answers a question well, it could mean Google expanded AI Overviews to more of your query set, or it could mean a competitor dropped out and you filled the gap by default. A falling trend has the same ambiguity in reverse, and it can also mean nothing changed on your end at all, since Google's own criteria for when AI Overviews trigger shift constantly and independent of your content. Read the trend as a directional signal worth investigating, not a verdict on a specific page or a specific piece of work.

The new opt-out toggle, and why blocking AI features doesn't touch core search

Alongside the report, Google added a toggle that lets site owners block their content from AI Overviews, AI Mode, and AI Overviews in Discover (SE Roundtable). Both the report and the toggle rolled out first to a subset of website owners in the UK, with the opt-out taking effect on June 17, 2026 ahead of a wider release (Brodie Clark Consulting).

The part worth underlining: Google states the opt-out is not used as a ranking signal for standard web search results. Turning it on removes your impressions and any downstream clicks from AI Overviews, AI Mode, and AI Overviews in Discover, and it leaves your organic rankings alone. For most publishers that makes the opt-out a narrow, deliberate choice rather than a default worth flipping, since the trade is giving up AI-surface visibility (and the citation premium Seer measured above) with no upside for the rest of your search presence. It's the kind of switch to consider case by case, not something to set once for an entire property.

Where this free report fits next to paid GEO tracking tools

Search Console's AI report is free where every dedicated GEO tracker charges monthly, and that's genuinely useful, but it answers a narrower question than any of them do. Think of it as the free first layer of a stack, not a replacement for the rest of it, which is the same framing we use in the fuller AI citation tracking guide.

What GSC gives you free that paid trackers charge for

A basic "are we appearing in AI search at all" signal, first-party from Google, with page, country, device, and date breakdowns, at no cost. That alone used to require a paid subscription or a manual prompt log to approximate. If you're only trying to answer whether your content shows up in Google's AI features specifically, the free report may be enough on its own for a while.

What it still can't do

Query-level detail, competitor share of voice, and click attribution stay firmly outside what GSC offers, and those are exactly what tools like Otterly, Profound, and Scrunch AI are built to provide. Our comparison of Profound, Otterly, and Scrunch AI covers pricing and engine coverage for all three in detail, and the broader roundup of AEO platforms is worth a look if you're comparing categories rather than three specific vendors. None of them, GSC included, will tell you which exact prompt triggered an impression on Google's side; that data isn't exposed to anyone yet.

There's also a scope limit worth naming directly: Search Console only ever covers Google's AI surfaces. It has nothing to say about ChatGPT, Perplexity, Claude, or Copilot, which is the gap a broader answer engine optimization measurement plan has to cover with other tools entirely.

A practical stack

Start with the free layer, since it costs nothing and is already in an account most teams have open weekly. Once your Generative AI tab clears the threshold and starts populating, use it to spot which pages and countries are picking up AI visibility on Google specifically. Layer a manual prompt log or a paid tracker on top for the surfaces and metrics GSC can't reach: cross-engine coverage beyond Google, query-level detail, and anything resembling a click or conversion number. GSC tells you where to look. A tracker or a prompt log tells you what happened when someone actually asked.

This is also why we built Lyra to fact-check every claim and attach real sources before a post ships: if she's writing the pages that end up in this report, you want to already trust the sourcing on them, not find out after the fact that a stat was wrong. Lyra's plans run on your own Anthropic key at cost, so measurement tools like the ones above are the only recurring bill in this stack besides Lyra herself.

A five-minute setup checklist

1. Open Search Console, go to Performance, and check for a Generative AI tab next to Search results.
2. If it's missing, confirm your property is large enough to clear the impression threshold and check back in a few weeks; there's no manual request to speed this up.
3. Once it's live, filter by page to find which URLs are picking up AI impressions, and cross-reference them against your highest-traffic organic pages to spot mismatches. The interesting cases are the two extremes: a page that ranks well organically but barely shows up here (Google's AI layer isn't drawing on it, even though people find it), and a page with heavy AI impressions but middling organic rank (it's answering the question well enough to get pulled into a summary without ranking for the query outright).
4. Check the country and device breakdowns for surprises, a spike in one country you don't otherwise rank well in is often the first sign a specific page answers a question especially clearly, in a way a plain keyword ranking wouldn't have surfaced.
5. Log the current impression trend as a baseline, then revisit monthly. Don't act on a single week of data; treat one-off spikes as noise until a trend holds for at least a full month.
6. Pair it with a lightweight manual prompt log, or a paid tracker if AI search already matters enough to your traffic to justify one, so you have a click and query signal this report doesn't give you yet.

Search Console's AI report tells you whether Google's answers include your pages; Lyra writes the kind of pages that earn that inclusion in the first place, fact-checked and shipped as a pull request you review before anything goes live.

Try Lyra → · Talk to the founder