As publishers await click data, Google’s AI visibility push complicates the opt-out decision

The opt-out is a Search Console control that lets publishers exclude their content from Google’s AI features. The control follows the CMA’s first conduct requirement on Google, imposed over the summer. While the ruling was U.K.-only, Google opened it to sites worldwide on August 31. Page-level grounding controls are due by March 3, 2027. 

But the controls are bundled: publishers can’t opt out of individual features. That leaves them weighing the potential visibility against the loss of control and waiting for click data to help judge the trade-off. 

That click data, which Google is required to provide to U.K. publishers by Dec. 3, will give publishers a key missing piece of the puzzle. And given Google made the first requirements global, publishers hope it will provide click data globally too. Google hasn’t confirmed that. 

Google has shown regulators and publishers that it is cooperating by sharing impression data: how often publisher content appears in AI features. But it has held back the click data, which would show whether those appearances actually send people to publishers’ sites. “That is the metric that will determine whether they [publishers] turn it off or not,” said Chris Dicker, CEO of Candr Media, which owns Trusted Reviews. “They’re not giving the metrics [currently] that actually mean anything, when they could have — because they’ve got those — they could have done it at the same time as impressions, but they haven’t.” 

He added that “everyone is holding off” playing around with the new controls until they see the click data because they don’t want to turn off a large proportion of their activities.

Dicker expects Google to share the click data in the coming weeks, ahead of Black Friday, when publishers are unlikely to risk experimenting with opt-outs during a crucial sales period. 

Meanwhile, Google has updated AI Mode and AI Overviews to show links more often and more prominently, added timely article carousels, and included links to related stories and more context around surfaced articles. 

Google says these changes will help publishers better track, control and increase their content visibility. It’s also testing links from video creators’ and publishers’ names to their Search profiles, and says that its Preferred Sources — the feature that lets users pick and bookmark the specific websites and publishers they most trust — has grown sevenfold this year. 

That visibility comes with a catch, though: the more valuable Google’s AI features become, the harder it may be to justify opting out. 

For some publishers, that trade-off is already a live question. “We’re discussing the CMA controls and how we exercise those,” said Carly Steven, director of ecommerce and commercial editorial at the Daily Mail, at Digiday’s Media Strategies London event in October. 

“We need to be so, so careful because just as we’re about to get the ability to remove articles on a URL basis, Google announces this new feature, and we have to ask ourselves: do we really not want to be there? It does still matter that we’re on the surfaces where users are. So it’s a very complicated picture.” 

David Buttle, founder of publisher AI standards coalition SPUR, believes Google’s preferred approach is to provide enough referral traffic from its AI-related surfaces for publishers to agree to let Google use their content, rather than paying them each time it does so. That means it needs to continue providing some kind of value exchange. “The way they can achieve that within the context of the conduct requirement is to ensure they can deliver enough value in the form of traffic from those surfaces back to publishers, so that they can say yes voluntarily,” said Buttle. 

Trying to maintain its exchange with publishers is tricky when its newest rival for information retrieval – OpenAI – isn’t constrained by the same legacy relationship. “ChatGPT doesn’t pay for content generally, and it doesn’t send any traffic out,” he added. “So Google’s approach is to try and intermediate as many clicks from Discover, which is a really important traffic source for publishers, and they’re serving summaries really aggressively in Discover, and bundling the control for that with the control for summaries everywhere else.” 

For all the noise around opting out, publishers are unlikely to do so unless they can opt out in a more granular way, stressed Buttle. “What it [Google] can do, though, is play around with the interface and the product for Discover to juice the AI-intermediated referrals that come out of that to get as much traffic out of that, so if you’re a publisher, it makes that decision to opt out far harder,” he added.

For some publishers, the question is still how to assess the controls before deciding whether to use them. At The Economist Group, that process is only beginning.

“On the CMA point, we’re looking at it,” said Josh Muncke, vp of generative AI at The Economist Group on stage at the Digiday Media Strategies event earlier this month. “I don’t believe we’ve tested it yet, but we’re certainly looking at it.”

Whatever that decision is, it will ultimately hinge on two things: one is how much traffic Google sends The Economist. The other is how often its name appears in the AI answers at the top of the results, even when nobody clicks through. He added: There are second-order effects to turning these things off that aren’t just about the traffic. They’re about broader visibility in the world.”

Some of those effects are already on the table. Some publishers expect excluded sites to drop out of the Top Stories carousel Google now places inside AI Overviews and Google hasn’t said how its control treats that unit. 

Reuters Digital Phil Andraos told Digiday in June that the lack of click data meant any decision could be slow and probably involve testing.

Muncke’s instinct runs the same way. He said he wouldn’t frame it as a single in-or-out call at a fixed point. “It will be a case more of thinking about where and how we can test, measure, and see the impacts, and make kind of like more structured, data-driven calls as we go,” he said.

The Professional Publishers Association (PPA) is appealing to the CMA to press Google to unbundle the controls so they apply to individual features, rather than solely site- and page-level, so that publishers can have a deeper level of control.

For example, a specific query answered by an AI Overview may replace a click a publisher would previously have received, while the same content could still be valuable in a broader, conversational AI search. 

“You’re either opting out of search AI Mode, AI Overviews – basically all these different features, or none of them,” said PPA’s head of policy and Public Affairs Eilidh Wilson. “And this is something that we continue to hear is what most concerns publishers. We’ve been really wanting to land this with impact with the CMA, so we’ve agreed to do a further follow-up briefing with them on it to look at the different parts,” she added.