Something unusual happened in UK tech policy this week: a regulator told Google to stop, and Google said okay.

The UK’s Competition and Markets Authority (CMA) has ordered Google to give publishers real tools to opt their content out of AI search features — AI Overviews, AI Mode, and the like — and to display proper links and attribution when their content is used. Google has nine months to comply. The company says it will.

This matters for a simple reason: it’s never really happened before. Nowhere else in the world has a competition regulator forced a search engine to give publishers control over how their content is used to power AI-generated answers. The EU has talked about transparency. The US has had hand-wringing hearings. But the UK’s CMA actually acted — and Google didn’t fight it all the way to court.

Why Now?

Publishers have been screaming about AI Overviews for two years. Their traffic has crater as Google summarizes their articles directly in search results, giving users the answer without a click. The argument has always been the same: Google is using our content to build a product that replaces our business model, and we can’t do anything about it.

That last part was the crux. Publishers could technically block AI crawlers via robots.txt, but that also blocked them from regular search — a trade-off no publisher was willing to make. Google held all the cards.

The CMA changed the geometry of that negotiation by ruling that Google cannot penalize publishers for opting out of AI features. You can exclude your content from AI Overviews and still rank normally in search. That’s the part Google fought hardest against in its February submission to the CMA, arguing — with a straight face — that “excessive attribution of lots of sources may worsen the user experience.” The CMA didn’t buy it.

Nine Months Is a Long Time. But Also Not.

Google will have to implement the opt-out tools within nine months, though the CMA expects “important parts” to arrive well before that. Google’s own announcement of new Search Console controls suggests they’re already moving — they’re rolling out globally, not just in the UK, which is interesting. Once you build the tool, you ship it everywhere.

The attribution piece is the subtler fight. Google’s position has essentially been: we know how to balance attribution and usability, and you should trust us. The CMA’s response: no, you clearly don’t, because there are documented cases where attribution in AI Overviews has been inaccurate. Now Google has to publish how it ensures and measures the factuality of AI-generated search results.

Is This Good for the Web?

Here’s where it gets complicated. The CMA’s decision is good for publishers — finally, some leverage. But it’s worth being clear-eyed about what it doesn’t fix.

The underlying problem with AI Overviews isn’t really about attribution. It’s that confident-sounding answers can be generated from content that doesn’t actually support them, presented alongside links that may or may not substantiate the claim. Clearer attribution helps. It doesn’t solve the deeper issue of AI systems that can generate plausible-sounding nonsense with a thin veneer of sources.

That’s a harder problem — one that no regulator has figured out how to solve with an opt-out toggle.

What the CMA has done is something more modest but real: it has made Google’s AI search product a little more transparent, and given publishers a seat at a table Google had been reserving exclusively for itself. In a world where AI search is becoming the default interface for the web, that matters. It’s not a fix. But it’s a start.

Source: Ars Technica