MartechARTICLE

Google Search Console rolls out AI performance report and AI Overviews block control to everyone

The AI response performance report and the button to opt out of generative search have reached all Search Console accounts. Here's what you can measure, what's left out, and the trade-off of turning it off.

Google Search Console rolls out AI performance report and AI Overviews block control to everyone
Image: Sabrina Santos

Google completed the global rollout of two Search Console features that had been in slow release for months: the AI performance report and the control to block content from appearing in generative search. According to Search Engine Land, the company updated its documentation on August 31, 2026, noting that the control now applies "to all sites worldwide." In other words: any Search Console account in Brazil should already have access, without waiting for early access.

For those working in growth and organic media, this puts an old complaint to rest. Ever since Google launched AI Overviews and AI Mode, publishers had asked for two things: to know how much traffic AI generates and to be able to decide whether they want to take part in it. Both now exist, but each comes with a catch worth understanding before making a decision.

What the AI performance report shows (and what it hides)

The new report measures how content appears across three generative surfaces of Google Search: AI responses, AI Mode, and AI Overviews. The available data is the usual Search Console set (impressions, pages, countries, devices, and dates), with one critical omission:

The report includes impressions, pages, countries, devices, and dates, but does not include click data.

>

-- Search Engine Land, on the AI performance report

This gap is not a minor detail. Without clicks, there's no way to calculate CTR or attribute direct sessions to a citation in an AI Overview. What's left is appearance volume: you know AI is using (or showing) your pages, in which markets and on which devices, but not what the user did afterward. In practice, this means the report works more as a presence gauge than as a conversion metric, and that's exactly where today's GEO challenge lies: AI answers right on the SERP itself and the click disappears, so measurement needs to shift from "how much traffic am I pulling" to "how much am I appearing in the answer."

The sensible path here is to cross-reference the AI report's impressions with traditional search performance over the same period. If AI impressions rise while organic sessions in GA4 fall, you have a reasonable signal of cannibalization by AI Overview, something that used to be just a suspicion and now has at least one side of the equation backed by data.

The toggle that takes you out of generative AI

The second feature is a new button in Search Console that lets a site owner prevent their content from appearing and being used as a ground for Google's generative responses, including AI Overviews, AI Mode, and AI Overviews in Discover. In Google's words, the site owner "can decide whether they want their site to appear and help ground answers in Search's generative AI features."

This is where the update's trickiest trade-off comes in. Google was explicit on two points:

| Decision | Consequence | |---|---| | Opting out | The site receives no traffic or impressions from generative AI surfaces | | Continuing to participate | Content can be cited and used to ground AI responses |

The point that keeps panic at bay: according to Google, this control is not used as a ranking signal for traditional search. In other words, turning off AI does not demote your pages in classic organic search (so-called core web search). It's a clean separation between the two worlds, and that matters because the temptation would be to assume that "opting out of AI" would penalize the site across the board. That's not the case.

Even so, opting out is not a trivial decision. Leaving means giving up any visibility on the surfaces that are absorbing an ever-growing share of informational queries. For a media site that lives off pageviews, it may make sense to test the block and measure whether overall organic traffic recovers, since AI would be "summarizing" the content without returning the click. For a brand that wants to be cited as a trusted source within AI responses (branding, authority, GEO presence), opting out is shooting yourself in the foot.

How this relates to robots and noindex

It's worth understanding what this toggle replaces and what it doesn't replace. Google already offered Google-Extended in robots.txt to control content use in model training and in some generative features. The new Search Console control is more direct and granular for Search surfaces: a visual switch, with no need to edit a file and no risk of getting the robots syntax wrong.

The layered difference looks like this:

  • noindex: removes the page from the index entirely, disappearing from both traditional search and AI. Too radical for those who only want to opt out of AI.
  • Google-Extended in robots.txt: a directive-level control, historically tied to training and grounding, with less transparency about its effect.
  • Toggle in Search Console (new): a specific opt-out from Search's generative features, without affecting organic ranking, with an effect documented by Google.

For the Brazilian team, the practical recommendation is not to go changing everything at once. The toggle is the surgical tool; noindex remains the atomic bomb.

What to do with this in practice

The real value of these tools shows up when you measure before and after. A roadmap that makes sense for an average Brazilian site:

  1. Open the AI report and establish a baseline: note AI impressions per page over the last 90 days, segmented by country (focus on Brazil) and device.
  2. Cross-reference with traditional search: compare the classic organic impressions curve with the AI one over the same period. Where AI grows and organic clicks fall, you identify content being "consumed" on the SERP.
  3. Decide by section, not by the whole site: since opt-out is per property, whoever has high commercial-value content (product pages, comparisons, conversion guides) needs to think differently than someone with purely informational content.
  4. Document the hypothesis before turning it off: if you opt out, record the date and track for 30 to 60 days whether total organic traffic responds. Without this control, you won't know if the decision paid off.

When it's not worth touching

If your site depends on appearing as an authority source and the strategy is GEO, blocking AI is counterproductive: you disappear precisely from the surfaces capturing search intent. And if you have no way to measure total traffic with a minimum of rigor (GA4 set up, period comparisons), opting out becomes a bet in the dark, because Google's own promise is that you'll lose AI impressions and traffic with no guarantee of recovering it elsewhere.

The takeaway is less about the button and more about the shift in yardstick. Measuring search presence is no longer just clicks and average position: it now includes "how much I appear in AI responses," a metric Google finally exposes, even if only halfway, without the clicks. For those deciding channel budgets, this is the missing piece needed to start treating GEO as engineering, rather than guesswork.

Translated from the Brazilian Portuguese original · Read the original