
Google lets sites opt out of AI search results, knowing most have nowhere else to go
Quick Answer
Google has introduced an opt-out feature for website operators in Search Console, allowing them to exclude their sites from AI search results like AI Overviews and AI Mode, which collectively attract over 3.5 billion users monthly.
Quick Take
This decision follows concerns from the UK's Competition and Markets Authority regarding the disadvantages faced by website operators.
Key Points
- Website operators can now opt out of AI search features in Search Console.
- AI Overviews and AI Mode reach over 3.5 billion users monthly.
- The UK's Competition and Markets Authority prompted this opt-out option.
- New performance reports will separate impressions for opted-out sites.
- Most website operators may have limited alternatives to Google's search.
📖 Reader Mode
~3 min readGoogle is rolling out new tools that give website operators more control over how their content shows up and how they can measure it in AI-powered search features, likely in response to regulatory pressure.
Google says AI Overviews now has more than 2.5 billion monthly active users, with AI Mode topping one billion. Going forward, a new toggle in Google Search Console will let site operators choose whether their content appears in AI search features like AI Overviews or AI Mode.
Sites that opt out won't get any traffic from these features (which shouldn't be much of a loss), but it sounds like they'll keep getting normal search traffic. "This control will not be used as a ranking signal for search results outside of these generative AI Search features," the company writes. Google is also shipping new performance reports that break out impressions, pages, countries, and devices for AI features separately.

Both features will initially be tested in the UK, partly due to pressure from the UK Competition and Markets Authority (CMA), which also issued a first-of-its-kind code of conduct targeting Google today. Under the new rules, Google must let publishers remove their content from AI search features, clearly attribute sources in AI-generated search results with links, and only fine-tune AI models on publisher content with their consent.
Publisher opt-outs won't slow down or weaken Google
Google takes big chunks of the web through its AI answers and sends little traffic back to the sites that made those answers possible. There's no fair deal here. A recent New York Times study found Google's AI answers are right more than 90 percent of the time. That still leaves millions of wrong answers every hour. But it's accurate enough to break people of the habit of clicking through to websites. Hardly anyone taps a source link in an AI answer.
Fig-leaf features like "Preferred Sources" won't fix that. Publishers who play along are slowly bargaining away whatever leverage they have left. Google owns the platform, and publishers get two options: show up without fair pay, or disappear. Either way, the value flows to Google.
When top publishers opt out of AI features, they face a potential loss of visibility. For Google, it's a non-issue. It has plenty of other material to build answers from: Reddit (which cites publishers, and where Google is a major investor), Wikipedia (which cites publishers), forums (which cite publishers), SEO pages (which rephrase or cite publishers), and its own data (trained on publishers). Answer quality might slip a bit, but not enough to change how most people use search.
So what would actually slow Google down? A legal obligation to pay for publisher content used in AI answers, something like the EU's ancillary copyright for publishers, but with real teeth. Or an antitrust ruling that forces Google to split AI answers from traditional search results.
— Originally published at the-decoder.com
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from The Decoder
See more →
An AI model programmed nonstop for 19 days on a single MirrorCode task that cost $2,600 to run
Epoch AI's MirrorCode benchmark reveals Claude Opus 4.7 as the leader with a 56% solve rate, reconstructing a 16,000-line toolkit in 14 hours. Despite this, all models tested struggle with the most complex tasks, highlighting limitations in current AI capabilities. The single task consumed $2,600 over 19 days, raising questions about cost-effectiveness in AI development.

