Google’s forced AI opt out: what changes — and what doesn’t — for publishers

Related Insights

Media

The Rundown: Google has drawn its AI payment lines — and publishers’ leverage is narrow Read More

Google has responded by saying it plans to play ball.

Building on its existing framework (namely its controls provided to date, including featured snippets and Google-Extended) it will work with the web ecosystem to address the CMA’s request. “We’re now exploring updates to our controls to let sites specifically opt out of search generative AI features,” said Ron Eden, Google’s principal for product management, in a statement. “Any new controls need to avoid breaking search in a way that leads to a fragmented or confusing experience for people. As AI increasingly becomes a core part of how people find information, any new controls also need to be simple and scalable for website owners,” he added.

It’s a small win for publishers: they finally get some say in how AI features use their work.

But while publishers have welcomed the proposals as a firm stepping stone, there remain unresolved questions.

A quick reminder of what the CMA’s new rules are: 

• Publisher opt out from AI features: publishers can prevent their content from being used in AI Overviews, AI Mode and other generative AI services, while still appearing in normal search results. 
• Opt out must be effective at both site-wide and page levels, and Google cannot penalize or downrank sites that choose to opt out.
• Fair ranking and no penalties: publishers that opt out can’t be downranked or displayed differently in Google’s general search results.
• Google must publish clear information about how publisher content is used in AI training, grounding and generative AI features. Like engagement metrics, impressions, and clickthrough rates, for content appearing in AI features, separately from general search. 
• Google must clearly and accurately attribute publisher sources in AI-generated summaries and explain how it does this.
• For Google to clarify the purpose, coverage and limitations of existing tools like Google Extended for AI content use.
• Google cannot acquire publisher content through third-party scrapers to bypass opt outs. 

But…

Publishers want a structural remedy vs. a behavioral one 

The CMA has not formally called for a full, structural separation of Google’s AI and search crawler, which is what would be at the top of publishers’ wish lists. While the CMA’s current proposals strongly emphasize opt outs, attribution, transparency and search ranking protection, they don’t yet include a binding rule that Google must maintain separate technical crawlers for AI versus traditional search crawling.

Related Insights

Bold Call

Bold Call: AI will rewrite publishers’ websites in 2026 Read More

Technically, Google’s existing tools let publishers block AI training, but in practice, those controls don’t fully shield traffic or revenue, which is why publishers keep calling for a true separation of AI and search crawlers.

Separating its crawler isn’t a technical issue; it’s about whether Google has the business incentive to do it, stressed Owen Meredith, CEO of News Media Association, which counts The Times, The Guardian, and The Daily Mail, among its members. 

“We’re skeptical about a remedy that relies on Google to separate data for AI Overviews versus search after it has been scraped — this is a behavioral remedy, whereas the cleanest solution would be a structural remedy, requiring Google to separate its crawlers for specific purposes,” he said. 

Google laid out a plan document in 2024 for how it could separate its crawler, to show how easily it could do it, but chose not to. “I think if Google actually wanted to do it, and it was in their interest, they could do it by tomorrow. It’s easy and straightforward and they don’t do it because it gives them a competitive advantage over OpenAI and others,” said Paul Bannister, CRO of Raptive.

Publishers and publisher associations are deep into reviewing the CMA’s proposals. U.S. publisher trade association Digital Content Next CEO Jason Kint said the new conduct requirements offer “long overdue” tools for reach choice, transparency and attribution, but also underlined the need for a more structural remedy. 

“Across Google’s numerous antitrust losses globally, a consistent theme emerged: Google’s data advantages enable it to leverage one monopoly to entrench another — mining content and user signals to favor products where it seeks more profit and to shut out competition,” said Kint. “In that light, the question isn’t just whether these behavioral changes are well intentioned, but whether they are truly enforceable and sufficient. Structural separation, maybe in this case of Google’s crawlers and the data they each collect, must remain firmly on the table.”

This doesn’t solve the ‘AI ate the click’ problem

Google argued to the CMA that using publisher content to train AI doesn’t harm sites, because the models learn how to process information rather than replace current content.

It claims that fine tuning AI with publisher content won’t substitute for websites, since static models quickly become outdated and prone to hallucinations. Therefore, AI training uses patterns from content but doesn’t create a live replacement for publisher pages — so publishers aren’t at risk. 

Naturally, publishers see that very differently. Even if the AI isn’t a perfect substitute, the use of their content in AI summaries and features siphons clicks, reduces ad impressions and cuts revenue — risks that the “no harm” argument largely bypasses. And some don’t trust Google not to find workarounds. 

Meanwhile, in theory, opt-out does give publishers greater leverage in negotiations with Google in terms of licensing content for grounding (Google has said it won’t ever pay for training), but the CMA has postponed a formal rule requiring Google to negotiate fair payments for at least 12 months. That has been a “disappointing” outcome, according to Meredith. “The government must ensure that the CMA is adequately resourced so it can move forward at pace with this, and other crucial interventions,” he said. 

Although opt‑out is an essential safeguard, it doesn’t resolve the wider value‑exchange question, stressed Sajeeda Merali, CEO of the Professional Publishers Association, which represents approximately 250 media organizations. “AI Overviews still replace clicks in many contexts, and without a clear model for licensing, the commercial imbalance remains,” said Merali. She said the proposal is a major step forward, but that “it is not the end of the conversation on publisher compensation.”

At this point, meaningful transparency about the content that’s being used is what publishers care about. “A problem we’ve seen across Europe is that Google generally doesn’t care about publishers exercising their opt out of and continues to use their content freely, even behind paywalls,” said Iacob Gammeltoft, senior policy manager at News Media Europe, which represents over 2,7000 European news brands. “So the opt out by itself is not enough and needs to be accompanied by a regulatory process that creates accountability.”

Others think that table stakes for transparent licensing terms should also be part of the CMA’s measures, so publishers don’t get lumped with unfair terms. “There is a transparency issue here. If Google is allowed to make deals with publishers for their content without the terms of licenses being disclosed, then the informational asymmetry gives Google a real advantage,” said Dr. Tim Squirrell, head of strategy for Foxglove, a tech justice organization that has lobbied the CMA on the issue.

Publishers want to see proof in the data

A common bugbear among publishers has been Google’s repeated stance that publishers are measuring the effect of AI Overviews on their referral traffic incorrectly or in an outdated way, without offering any data evidence to back that up.

Merali said publishers want proof, proper checks, and practical enforceability. “The wish list is simple: clear data on how AI affects visibility, reliable guarantees that opting out won’t disadvantage them, and an auditable system so we can independently verify what’s happening in practice,” she said.

While it’s early days, publishers do believe the regulatory scrutiny the CMA has kickstarted will put in some better guardrails. “The CMA has the power to monitor compliance and demand information from Google and has also committed to engaging with impacted parties, such as news publishers, to understand if the remedies are effective,” said Meredith. “Non compliance could result in penalties of up to 10 percent of Google’s global turnover. We will expect the CMA to act swiftly to implement more stringent requirements if they find evidence of non compliance,” he added. 

However, others remain unconvinced that it will be enough. “I expect them [Google] to delay and obfuscate this as much as they can, while claiming that they are working with the CMA,” said Bannister. “If the CMA comes down with a hard-and-fast rule, I also expect them to implement it in the most ham-fisted way so they can say they are compliant but do it in a way that still makes it really impossible.”