Google's AI search opt-out went live June 6, 2026, with a Generative AI report showing impressions but no clicks. What the CMA required and how to decide.
nosnippet, max-snippet, and data-nosnippet.The Google AI search opt-out that publishers spent two years asking for is now a real switch. On June 6, 2026, Google added a toggle to Search Console that removes a site from AI Overviews, AI Mode, and AI features in Discover, currently live for UK site owners with a stated plan to expand globally. The same week, a new Generative AI report appeared under the Performance tab. Real progress, both of them. The catch is what the report leaves out: it shows impressions only. No clicks, no click-through rate, no way to know whether your AI Overview appearances send you anything at all. Google handed publishers the steering wheel and kept the speedometer.
That gap should shape how you use the toggle. Here is what the UK regulator actually required, what Google shipped against it, and how to make the opt-out decision with data instead of sentiment.
This started as a regulatory order, not a Google initiative. The UK Competition and Markets Authority designated Google with "strategic market status" under the Digital Markets, Competition and Consumers Act, a designation the CMA was careful to note is not a finding that Google broke competition law. It simply means Google's position in search is entrenched enough that the regulator can impose conduct requirements directly.
The CMA imposed three:
The timeline is staged. The core conduct requirement took effect June 6, 2026, roughly six months after the CMA published its decision. Additional obligations land in December 2026, and Google has until March 2027 to deliver page-level controls. Google must also file compliance reports with the CMA every six months for the first year, and the regulator has signaled that further announcements about Google's search business are coming.
CMA chief executive Sarah Cardell framed the stakes plainly: with features like AI Overviews rapidly reshaping search, publishers need "appropriate bargaining power over how their content is used." Strictly speaking, all of this applies to UK search users only. In practice, it is the template every other regulator will read first.
Google met the deadline on day one. The toggle went live in Search Console on June 6, 2026, the same day the conduct requirement took effect, and it covers three surfaces: AI Overviews, AI Mode, and AI Overviews appearing in Discover.
Three things are worth being precise about, because the coverage has blurred them.
First, it is currently available to UK site owners. Google says global expansion is planned, but if you manage a US-only property, you may not see the control yet. Second, it operates at the domain level. There is no way to keep your product pages in AI Overviews while pulling your editorial content out; it is all or nothing until the page-level controls arrive by the March 2027 deadline. Third, Google has confirmed that using the opt-out will not be punished with ranking adjustments in standard search. Your blue links stay where they were.
That last point matters more than it sounds. Until now, every mechanism for limiting AI reuse of your content carried a search-visibility cost. This is the first one that doesn't, at least on Google's word.
Alongside the toggle, Search Console gained a dedicated "Generative AI" section under the Performance tab. It shows how often your pages appear across AI Overviews, AI Mode, and AI features in Discover, and you can filter by page, country, device, and date range. Like the toggle, it is rolling out UK-first.
What it does not show: clicks, click-through rate, or position. Impressions only.
That is not what the regulator asked for. The CMA's interpretive notes specify that Google should provide impressions and engagement data, including click-throughs and CTR, via Search Console. Google shipped half of it and has not committed to a date for the other half. Search analyst Glenn Gabe summed up the practitioner reaction in one line: "AI reporting coming to GSC! Awesome! No click data. NOT Awesome."
The practical consequence is brutal in its simplicity. The toggle asks you to make a trade: give up AI-surface visibility in exchange for not having your content summarized. The report tells you the size of the visibility, but not its value. You can see that a page earned 40,000 AI Overview impressions last month. You cannot see whether those impressions produced 4,000 visits or 4. The single number you need to price the trade is the number Google withheld.
Before June 6, the only way to keep content out of AI Overviews was the snippet-control toolkit: nosnippet, max-snippet, and data-nosnippet. They worked, technically. They also gutted your normal search results, because the same directives that stop Google from quoting you in an AI answer stop it from showing a description under your blue link. Sites that tried it watched organic CTR drop on every query, AI-related or not.
It was a blunt instrument with collateral damage built in, which is why so few publishers used it despite loud frustration with AI Overviews. The new toggle is the first control that separates the two decisions: you can stay fully present in classic search, snippets intact, while exiting the AI layer. As a piece of mechanism design, that is genuine progress, and Google deserves the credit for shipping it on the compliance deadline rather than after it.
Note what the toggle is not, though. It is not Google-Extended, the robots.txt token that addresses Gemini model training and grounding. The Search Console toggle governs where your content appears; the training opt-out the CMA required is a separate obligation, and blocking crawlers is a separate decision again. If you are weighing the crawler question, our GPTBot decision guide walks through the same trade-offs on the OpenAI side.
So should you flip it? Not yet, and not because the toggle is bad. Because you almost certainly cannot price the decision.
Consider what each choice costs. Staying in means your content keeps feeding AI answers that may satisfy users without a click, the zero-click problem that drove the CMA case in the first place. Opting out means vanishing from a surface that, by Google's own positioning, sits above the classic results for a growing share of queries. If AI Overview citations were sending you meaningful traffic, you just cut it. If they were sending you nothing, you lost nothing and reclaimed some bargaining power.
Which scenario describes your site? The Generative AI report cannot tell you, because the answer lives in click data it doesn't have. Standard GSC reporting cannot tell you either, since clicks from AI Overviews are folded into ordinary organic totals with no separate label. And the toggle is domain-wide, so a wrong guess applies to every page you own at once.
There is also an asymmetry worth respecting: opting out is reversible, but the measurement window isn't. Once you flip the switch, your "before" data stops accruing. Whatever baseline you have on that day is the baseline you keep.
The data Google withheld is the demand side: what users do with your content inside AI surfaces. But there is a supply side you control completely, and it lives in your server logs: which AI crawlers fetch your pages, which pages they want, and how often they come back.
Put the two together and the fog clears considerably. A workable picture has three layers:
This layered approach is the same logic behind a default-deny posture for AI crawlers: you cannot set policy on traffic you have never measured. The opt-out toggle just raised the price of guessing.
Days 1 to 14: capture your baseline. If you have UK traffic, open the Generative AI report and export impressions by page and country before you change anything. Start logging AI crawler activity server-side the same week. Every day you wait is baseline you lose.
Days 15 to 45: segment your exposure. Rank pages by AI-surface impressions, then overlay organic clicks. Sort your content into three buckets: pages that thrive in both, pages with AI visibility but starving organic clicks, and pages AI surfaces ignore. The middle bucket is where the opt-out argument is strongest.
Days 46 to 75: decide, with a tripwire. If your high-value content sits mostly in that middle bucket and AI crawl volume keeps climbing, opting out is defensible; annotate the flip date and watch organic clicks and impressions for at least four weeks, including Discover. If your top pages win in both worlds, staying in and competing for attributed citations is the better trade. Either way, write down the metric that would reverse your decision before you make it.
Days 76 to 90: prepare for page-level controls. The March 2027 deadline means granular opt-outs are coming. The sites that benefit will be the ones holding per-page AI data when the controls arrive. Build that dataset now and the next decision is a sort, not a debate.
Google shipped a real control and half the data needed to use it. The other half is sitting in your logs.
Written by Crawlytics Team. Crawlytics tracks AI bots, generates llms.txt, and powers WebMCP commerce, all from one snippet on any stack. See how it works →
No. Google has confirmed that using the Search Console opt-out toggle will not trigger ranking adjustments in standard search results. Your pages remain in the classic index with snippets intact; they simply stop being used to generate AI Overviews, AI Mode answers, and AI features in Discover. This is the key difference from the old nosnippet approach, which suppressed your descriptions in regular results as a side effect. That said, AI surfaces increasingly sit above classic results, so opting out trades AI-layer visibility for content control even though your rankings stay put.
For now, yes, in availability. The toggle and the Generative AI report went live for UK site owners on June 6, 2026, because they implement the UK CMA's conduct requirements under Google's strategic market status designation. Google has said global expansion is planned but has not published a date. The CMA's obligations also phase in over time: additional requirements take effect in December 2026, and page-level controls are due by March 2027. Publishers outside the UK should treat the UK rollout as a preview of the controls they will likely get, and start baselining their data now.
No, the Search Console toggle controls where your content appears, not what models train on it. The CMA imposed a separate training opt-out obligation, which it called a world first, requiring Google to let publishers exclude their content from AI model training. The appearance toggle that shipped on June 6 addresses the AI feature obligation. If training is your concern today, the existing mechanism is the Google-Extended robots.txt token, which governs use of your content for Gemini training and grounding. The two controls are independent: you can opt out of one, both, or neither.
They operate at different layers. Blocking Google-Extended in robots.txt tells Google not to use your crawled content for Gemini model training and grounding, but it never removed you from AI Overviews, which are built on the standard Googlebot crawl. The Search Console toggle does the opposite: it removes your content from AI Overviews, AI Mode, and AI features in Discover, while saying nothing about training. A publisher who wants maximum control would use both, plus crawl-side monitoring to verify bot behavior actually changes after the directives go in.
It appears as a dedicated "Generative AI" section under the Performance tab in Search Console, currently rolling out to UK properties first. The report shows impressions across AI Overviews, AI Mode, and AI features in Discover, filterable by page, country, device, and date range. It does not include clicks, CTR, or position data, although the CMA's interpretive notes say engagement data should be provided. Until click data arrives, the most useful move is exporting impressions by page and cross-referencing them against your standard organic performance and server-side AI crawler logs.
This page is part of Crawlytics.app. View all pages: llms.txt · llms-full.txt