Should You Opt Out of Google AI Search? The UK Mandate Explained

For the first time, website owners are getting a "refuse AI search" button. The question: should you press it?

In June 2026, the UK's Competition and Markets Authority (CMA) imposed a world-first requirement on Google: publishers will be able to opt out of having their content power AI features in Google search, and opt out of their content being used to fine-tune AI models (CMA official announcement, 2026). UK media celebrated; coverage spread worldwide; and site owners everywhere started asking the same thing: when this button reaches me, press or not?

This article explains the event first, then answers the bigger question. Conclusion up front: for the vast majority of small businesses, the answer is don't opt out — but the reasoning is worth reading.

💡 Not sure whether AI search is a threat or an opportunity for your site? Free consultation

Key takeaway: The UK CMA requires Google to let publishers opt out of both "content powering search AI features" and "model fine-tuning," with clear source attribution in AI results, fully implemented within nine months (CMA official, 2026). A world first — but opting out costs most SMBs more than it protects.

What Did the UK Demand From Google?

The official terms, not the retellings. After designating Google with strategic market status (SMS) in general search, the CMA required (CMA official, 2026):

Opt-out from AI features: publishers can prevent their content powering AI features like AI Overviews
Opt-out from fine-tuning: publishers can refuse content being used to fine-tune AI models
Attribution: Google must properly attribute publisher content with clear links in AI-generated results
Timeline: all changes within nine months, important controls expected sooner
Accountability: compliance reports with key data every six months

What the announcement didn't spell out but Google has since shipped: the control lives in Search Console under Settings → Search generative AI, set per property as Include / Exclude / Inherit from parent, and it covers AI Overviews, AI Mode, and AI Overviews in Discover; Google also states the setting is not used as a ranking signal (9to5Google, 2026). The "page-level control" mentioned in press coverage has no confirmed official implementation yet, so treat it as reporting rather than fact. Some coverage also cites a very high Google search share in the UK, but we couldn't find a citable official source for that percentage, so we don't quote a figure here.

Two easily-overstated points. First, the CMA explicitly says SMS designation "does not imply that it has acted anti-competitively" — this is ex-ante regulation, not a guilty verdict. Second, the first wave of scope is the UK: Google built the control into Search Console in June 2026, opened it to some UK sites first, and has publicly said it is rolling it out to website owners globally (9to5Google, 2026). Most properties elsewhere haven't been reached yet. The second half of this article is the decision framework for when it lands in your account.

What Exactly Are You Opting Out Of? Two Tiers

Retellings flatten "opt out" into one action; the CMA actually mandates two independent choices of very different natures.

Tier	What you opt out of	Effect
Tier 1: AI feature use	Content not used in real-time answers like AI Overviews	You vanish from AI answer citations
Tier 2: Model fine-tuning	Content not used to fine-tune AI models	Your content stays out of training updates

Look familiar? It's the same logic as the robots.txt world's "search crawlers vs training crawlers" — as covered in our AI crawler setup guide: allowing search and managing training are two separable decisions.

The difference is force. robots.txt is a gentleman's agreement (and 30% of AI crawler scrapes ignore it in practice); the CMA requirement is a legal compliance obligation — Google must build controls that actually work. That's what "world first" means here: a regulator turned choice into hard spec.

Why Do Publishers Want Out? Read the News Media's Ledger

To judge your own case, understand what the people who want out are calculating.

UK news media's model: content attracts clicks → clicks monetize (ads, subscriptions). AI Overviews serve the answer directly, intercepting the click on Google's page — content used, traffic gone. The zero-click data backs the anxiety: 68% of US Google searches ended without a click in Jan–Apr 2026, and CTR drops nearly 60% when AI Overviews appear (Search Engine Journal, 2026).

Media also have a second monetization path: licensing content to AI companies. The opt-out is negotiating leverage — "pay, or we pull out." Leverage works because the exit threat is now credible: before, Google could ignore them because media had nowhere to go; with regulation paving the exit, the table tilts.

So publishers fighting for opt-out are fighting for copyright monetization leverage, not expressing hatred of AI. It's a different front of the same war as the news-site crawler blockade (79% of top news publishers block at least one AI training bot).

Should Your Site Follow? Three Site Types, Three Answers

Back to your website. Split the question into three scenarios and the answer surfaces itself.

Scenario 1: You Monetize Content Directly (News Media, Paid Content)

Seriously evaluate opting out (when the option exists). Your ledger matches the UK media's: content is the product; AI taking it free erodes the core business. Opting out while keeping licensing-negotiation room is a coherent strategy.

Scenario 2: You're a Content-Marketing-Driven SMB Site (Most Readers Are Here)

Don't opt out — double down the other way. A completely different ledger: your content isn't the product, it's the customer-acquisition channel — articles exist so potential clients find you. AI citing your content is free exposure plus a trust signal; opting out is delisting yourself from the fastest-growing new channel and gifting the space to competitors who stayed.

Every client we serve falls in this category, and the play is always "make AI cite you correctly" rather than "hide from AI": allow search crawlers, strengthen structure, build measurement. The few clicks an SMB loses buy a position inside AI answers — currently a good trade.

"Currently" isn't "forever," so here are the re-evaluation triggers. Re-run the math when any of these appears:

Measured net loss: two consecutive quarters of rising AI exposure with clearly falling total site entries (including branded search), and no other explanation
Your content starts monetizing directly: subscriptions, paid courses, licensing — the ledger flips from "acquisition" to "product," and you move to Scenario 1
An equivalent mechanism reaches your market: by then you'll have GSC AI report history and manual test data — the button decision takes five minutes with your own numbers

No trigger, no monthly anxiety. Swap worry for a quarterly review and follow the data — emotions can track the headlines; decisions should track your own numbers.

Scenario 3: You Host Sensitive or Highly Proprietary Content

Partial opt-out (specific directories or pages) is worth considering, but check the granularity first: Google's Search generative AI control is set per property, and the official documentation lists no page-level option (Search Console Help, 2026). To fence off individual pages today you still need robots.txt plus page-level markup like nosnippet / data-nosnippet (Google AI features documentation). Either way, that's fencing off restricted zones, not full retreat.

Site type	Recommendation	Why
Media / paid content	Evaluate opt-out + license	Content is the product
Content-marketing site	Stay in, optimize harder	Content is the acquisition channel
Sensitive content	Fence off zones with robots.txt / `nosnippet`	Fence zones, don't retreat

Which Type Is Your Site?

If the call isn't obvious, let us take a look. AI SEO Hacker offers a free site health check including your current AI search exposure.

Free consultation

Staying In? Convert Exposure Into Assets

If you stay in AI search, maximize what staying buys you. Three moves:

Make sure AI gets in and reads you right: robots.txt allowing search crawlers (setup guide), self-contained paragraphs, key numbers up front
Measure the AI-side return: the Search Console AI performance report (launched June 2026) gives official exposure data; cross-platform citations via fixed-question tests (measurement guide)
Watch the attribution dividend: the CMA requires "clear links" crediting sources — the more visible attribution becomes, the more a citation converts into clicks and brand recognition

The attribution dividend changes the value structure of being cited, and it suggests three writing adjustments: put your brand name inside quotable paragraphs ("according to {brand}'s testing…") so attribution is built into the content; pair key numbers with sources and dates (freshness is citation competitiveness); and give every article a one-or-two-sentence quotable conclusion — that's the unit AI summaries lift. These work in unregulated markets too, because their essence is making citations carry your brand.

The bigger GEO-vs-SEO picture: see GEO vs SEO differences.

robots.txt: The "Half Opt-Out" You Already Have

The UK button hasn't reached most markets, but you already hold a rough version of these controls: robots.txt AI crawler management.

Item	CMA mechanism (UK)	robots.txt (available now)
Granularity	Property level, features vs fine-tuning split; no page-level option in the official docs	Per-crawler, per-directory
Force	Legal obligation with compliance reports	Honor system (30% of AI scrapes don't comply (Search Engine Journal, 2026))
Ranking after opt-out	Google states the control "isn't used as a ranking or inclusion signal affecting other parts of Search" (Search Console Help, 2026)	Blocking the wrong crawler can hurt indexing
Scope	UK publishers	Any website, worldwide

Our standard client configuration: search-purpose crawlers (OAI-SearchBot, Claude-SearchBot, PerplexityBot) all allowed; training crawlers (GPTBot, ClaudeBot) decided case by case — the civilian version of the CMA's two tiers. In other words, you can rehearse the "should I opt out" decision today with robots.txt, reversibly and for free. Per-crawler steps in the AI crawler setup guide.

Mind the granularity difference: robots.txt controls whether crawlers get in; the CMA mechanism controls what content gets used for. The former is the front door — close it and nothing else matters. That's why we usually advise keeping the door open and steering usage with content strategy.

Where Does This Regulation Go Next?

Announced facts only, no fortune-telling.

Nine-month implementation: all changes within nine months of June 2026, key controls expected earlier (CMA official, 2026)
Six-month compliance reports with key data and metrics
Reality has outrun the reported milestones: the Search Console control reached some UK sites in June 2026, well ahead of the "main publisher controls by December 2026" figure in press coverage (9to5Google, 2026); "page-level controls by March 2027" has no official source we could verify and remains press reporting only

For readers outside the UK, the point isn't the British timeline — it's the template effect: the world's first hard spec for AI-search choice now exists, ready for other jurisdictions to copy. What you should do now is figure out, with data, whether AI search is a net gain or net loss for you — so when the button arrives, your decision runs on evidence instead of headlines.

FAQ

Can sites outside the UK opt out of Google AI search now?

The mechanism now exists — it just hasn't reached most properties yet. Google shipped the control in June 2026 under Search Console → Settings → Search generative AI, set per property as Include / Exclude / Inherit, opened to some UK sites first, with a stated intent to roll it out globally (9to5Google, 2026). The CMA mandate itself still applies only to the UK, where Google has nine months to implement controls (CMA official, 2026). Practical step: check your Search Console settings page to see whether the option has appeared for your property. If it hasn't, the available tool is robots.txt-level crawler management — separately controlling search and training crawlers, as in our AI crawler setup guide.

Does opting out of AI search hurt Google rankings?

No. Google has explicitly said the control is not used as a ranking signal: after opting out your content still appears in regular Search and Discover, and Merchant Center and Google Ads are unaffected (9to5Google, 2026). What's certain: opting out removes you from AI Overviews citations, and AI features now appear on more than 20% of searches (Search Engine Journal, 2026) — a growing surface.

Why shouldn't small businesses opt out?

Different ledger. Media content is the product — AI summarizing it is taking the product free. SMB content is an acquisition channel — AI citing it is free exposure and trust. Opting out means delisting from the fastest-growing channel while competitors stay. The better play: optimize structure so AI cites you correctly, then measure the return via the GSC AI report and manual testing.

What's the difference between opting out of "fine-tuning" vs "AI features"?

Opting out of AI features keeps your content out of real-time answers like AI Overviews (affects current exposure); opting out of fine-tuning keeps it out of model training updates (affects long-term knowledge absorption). The CMA requires both choices separately (CMA official, 2026) — mirroring the search-crawler/training-crawler split in robots.txt.

I already block AI crawlers in robots.txt — same effect as the UK opt-out?

No. robots.txt is door-blocking — crawlers stay out, but 30% of AI scrapes ignore it (Search Engine Journal, 2026), and blocking the wrong search crawler can hurt regular indexing too. The UK mechanism is usage authorization — content gets crawled normally but you dictate what it may power, with legal accountability. Coarse vs fine; configure the coarse tool carefully.

Choice Is Good — Just Don't Rush to Use It on Yourself

Three sentences: the UK got the world's first hard-spec AI search choice, landing within nine months; those who want out are playing a copyright-monetization hand — that's the media ledger; the SMB ledger is acquisition — stay in, optimize, measure, and decide with your own data when the button ever knocks.

Regulation will keep moving and AI search will keep growing. Neither waits — but your data can be ready. The most practical next step is two actions: check your robots.txt crawler settings this week, and stand up AI exposure measurement this month. When a local version of the choice arrives, you'll be among the few who can decide in five minutes.