As Google search blue links face extinction, what does AI-first search mean for publishers?
BY HAZEL BROADLEY, BEELER.TECH
Somewhere inside a publisher’s analytics dashboard, a search visit that used to exist has quietly disappeared. Not because the story was weak, or the headline failed. But because a reader had a question and Google’s AI had the answer. The journey ended before the user ever left Google’s doorstep.
For years, publishers understood the search deal even when they didn’t love it. They made useful content, Google organized the web, users clicked through, and some of the resulting ad revenue found its way back to the source. The arrangement was never completely fair, but it was clear enough for publishers to build revenue strategies around it. Now, AI search is driving a harder bargain because the user no longer needs to even visit the publisher’s site to get an answer.
The blue link may not be vanishing overnight, but that’s exactly why the industry risks underestimating the change. While links may still be visible – neatly arranged around the AI’s answer – visibility is not the same as value. If the user no longer needs the link, it risks becoming irrelevant to the ecosystem which once made it powerful.
The old search habit is disappearing
The last 18 months have turned AI search from a product feature into a fight over distribution. OpenAI made ChatGPT search available to all logged-in users in December 2024, then pushed further into research workflows with “deep research” in February 2025. Microsoft added Copilot Search to Bing in April 2025, and Anthropic has been building web search and research into Claude.
AI search doesn’t behave like the old search results page. It breaks questions into smaller tasks, runs multiple searches, synthesizes what it finds, and often gives the user a usable answer before any click happens. Google calls this “query fan-out,” and its own Search Central documentation says that AI Overviews and AI Mode can use this approach to find supporting pages and links which may differ from classic search results.
Evidence of the blue link waning is strong:
- Pew Research found that Google users clicked a traditional result link in only 8% of visits when an AI summary appeared, compared with 15% of visits when one did not.
- Ahrefs, using a different methodology, found that AI Overviews were associated with a 58% lower average CTR for the top-ranking page.
But this is unlikely to affect every publisher to the same degree. After all, breaking news, specialist reporting, community content, and evergreen service journalism will not all behave in the same way. But the direction of travel is clear: fewer visits to the source page.
The commercial model is changing at the same time. Traditional search monetization was built around the page, the keyword, the click, and the ad unit. AI search adds a new cost structure, because every long answer, follow-up, agentic task, and generated interface consumes tokens and computing time.
That token economy has become impossible to ignore. Google said at I/O that developers are processing more than 19 billion tokens per minute across its APIs and tools, and that more than 375 Google Cloud customers have each consumed over one trillion tokens in the past 12 months. At the same time, corporate AI spending has started to evoke “sticker shock,” with Axios reporting that executives are questioning whether fast-rising AI costs are producing enough return.
Publishers are making a different point about cost though. If tokens aren’t free, then neither is the source material on which the AI relies, and which publishers produce. And this also encompasses the reporting, editing, photography, and fact-checking work that AI systems are trying to summarize. The web has always depended on someone paying for the work upstream, and AI search makes that dependence obvious, even as it weakens the path back to the people who funded the work.
What do we do about the news from Google I/O and the CMA?
Google’s I/O announcement in May was not a quiet product tweak. In its official recap, Google described a new era for AI Search, with a more intelligent search box, multimodal input, AI Mode follow-ups, background agents, agentic booking, shopping support, and custom generative interfaces.
Google framed the announcement around scale and user adoption, claiming that AI Overviews have reached more than 2.5 billion monthly users, and that AI Mode has passed one billion. The search giant also said that AI Mode queries had more than doubled every quarter since launch, which suggests that users are getting comfortable asking longer and more complex questions inside the search experience itself.
On the face of it, the direction of travel is worrying for publishers. AI Overview follow-ups can move users into AI Mode rather than down the page to blue links. Search agents can monitor topics in the background, generative UI can create custom layouts and tools on the fly, and commerce features can shorten the journey between query and transaction.
But there is a reprieve, at least for now. AI Mode is not the default search experience, standard links still exist, and Google has made updates designed to surface more links and source context in AI experiences. Google’s Search Central guidance also says there is no special optimization required for AI features, beyond being indexed and eligible to show a snippet in Search.
The CMA (the UK’s Competition and Markets Authority) then published a ruling just over two weeks later, on June 3rd, that directly addressed this move towards AI search. In a world-first, the regulator announced a conduct requirement which obligates Google to give publishers control over how their content is used in Search AI features. And, crucially, it said that publishers must be able to opt out of those uses without being penalized in standard Google Search ranking.
There are also other significant details. The CMA said that publishers must be able to opt out of having their content used to power Google’s Search AI features and model fine-tuning. It also required clear attribution, links to publishers, monitoring, implementation within nine months, and compliance reports every six months in the first year.
Google responded the same day with new tooling. It said it was testing a Search Console control for generative AI Search features, starting with a subset of UK site owners and then expanding globally. The new controls will let website owners decide whether their pages can appear in generative AI Search features, including AI Overviews and AI Mode, or be used as source material for those AI-generated answers. Google also said the opt-out will not be used to downgrade those sites in ordinary search results, but that those sites will also not receive traffic or impressions from AI features.
That last point is the new bargain in miniature. Publishers can protect their content from one use, but they may lose exposure in the very surfaces where user behavior is moving.
From fear to fatigue: publisher reactions to changes in search
It would be reductive to say publishers are “panicking.” There is anxiety, yes, but there is also fatigue, realism, and a large degree of nuance. Various teams are using AI, experimenting with it, building workflows around it, and finding genuine value, while still being deeply concerned about what AI search does to distribution.
Reactions to AI were captured well by the recent discussion around A.G. Sulzberger’s speech on AI, journalism, and the public square. Journalism and original publishing are not resistant to innovation; the objection is to uncompensated extraction and attribution, and letting one side capture the interface while the other carries much of the cost of producing trustworthy information.
The CMA ruling has therefore been welcomed, but not treated as a full answer. The News Media Association called it a “significant step towards levelling the playing field,” and the Publishers Association welcomed the ruling, but its CEO, Dan Conway, highlighted that this was the beginning of a difficult phase: “Now the hard work starts of monitoring and enforcing these new conduct requirements for Google.”
Aside from adaptation, some publishers are also frustrated that the industry keeps being asked to respond faster than it can reasonably do so. Some see the I/O announcement as confirmation of a direction Google had already been telegraphing. Others admit that after years of huge platform announcements – from privacy changes to cookie deprecation to Privacy Sandbox – it’s become tempting to tune out until the practical impact is unavoidable.
Collective power at platform scale
The opt-out needn’t be a dilemma – in fact, it could be leverage. Large publishers may be able to test page-level or directory-level approaches, withhold high-value content, build licensing pressure, and negotiate from a position of some force. Meanwhile, smaller publishers may have the same formal rights, but far less practical ability to turn those rights into revenue or protection.
That’s how a two-tier web emerges. The largest publishers negotiate, while the long tail gets a control panel.
The collective response is already forming. On June 3rd, the SPUR Coalition said it had welcomed 30 new members, and it describes its work as shaping the rules, standards, and infrastructure that allow publishers and AI platforms to do business. Publishers’ efforts could be stronger if they can coordinate and express their machine-readable rights consistently, but individual leverage won’t be enough. The answer layer is being built at platform scale, so publisher rights need to become legible at the same scale.
That’s why measurement is crucial.
While Google has announced dedicated Search Generative AI performance reports for its AI features – including impressions, pages, countries, devices, and dates – publishers need reporting that shows what AI Search is really doing to their work. Visibility data is a start, but it won’t answer the most important commercial question unless publishers can see whether AI exposure complements, replaces, or cannibalizes ordinary search visits.
Without that, the market will keep relying on scattered examples instead of usable evidence. That uncertainty suits platforms more than publishers. If nobody can see clearly when an AI answer is standing in for a publisher visit, nobody can properly argue for the value being lost.
Not all content is equal before AIs
A lot of the web was built to answer predictable search demand. Basic definitions. Simple explainers. Thin comparison pages. Pages written less because someone had something distinctive to say, and more to capitalise on the keywords.
AI search is very good at absorbing that layer. It can summarize it, repackage it, compare it, and flatten it into a usable answer. The more interchangeable the content, the easier it is for the answer layer to swallow.
However, that doesn’t mean only deep investigative journalism survives. Publishers just need to be honest about what’s genuinely hard to replace, such as original reporting, specialist analysis, firsthand expertise, interactive tools, and community knowledge. These are not magic shields, but they create reasons for a user to go beyond the AI interface.
The hard truth is that AI search punishes content strategies that were already too dependent on low-differentiation search capture. Many publishers built those strategies because the economics of search rewarded them. But the reward system is changing. A page that was valuable because it ranked may be much less valuable when the answer can extract the useful part and leave the rest behind.
The commercial question now is brutal and simple: what does the reader still need from the publisher after the machine has answered? If the answer is “not much,” then the strategy is in trouble.
Direct relationships are no longer optional
For years, publishers have talked about direct relationships as a strategic priority. The phrase has become so familiar that it risks losing potency, but AI search brings it to the fore again.
A direct relationship is more than just a newsletter signup, an app install, or a registration wall. It’s a way to keep the publisher visible when discovery surfaces become unstable. It’s the difference between relying on a platform to send a user back, and giving the user a reason to return without being sent.
This is crucial because AI search can change habits. If users become accustomed to asking an interface for answers, the publisher’s task becomes harder. The publication has to become something more than just a source – it needs to become a voice, a utility, or a community. Something entertaining. Something worth keeping. That’s why newsletters, podcasts, memberships, events, and specialist products need to be prioritized, not left to one side.
This is also where publishers need to separate AI adoption from AI dependency. Using AI inside the business can be sensible. Within workflows, it can help teams analyze data, speed up repetitive work, and remove operational drag. But there is a line. AI that strengthens the publisher’s own product is not the same as AI that trains the audience to need the publisher less. A cleaner dashboard is not necessarily better business.
Yes, we’re having to defend the open web… again
Return to that missing visit in the analytics dashboard. It is easy to treat it as a small unit of loss. One click. One session. One reader who did not arrive. But multiply that moment across millions of queries, and the scale of the problem becomes clearer. This isn’t only about lost traffic. It’s about a complete platform shift.
The blue link won’t disappear tomorrow. It will still sit there, neat and familiar, beneath an AI answer. But a species doesn’t vanish all at once. First, its habitat shrinks. Then its role in the ecosystem weakens. Then everyone notices that it’s gone.
The CMA ruling is a win, but control alone won’t save the economics of publishing. Collective leverage is a good start, but the industry still needs licensing models that make commercial sense. Before those models can work, publishers need evidence of what AI search is doing to their content. That means reporting which shows what is actually happening, content strategies built around defensible value, and direct audience relationships strong enough to survive a platform shift.
Google isn’t wrong that users want better answers. They do. AI search can be genuinely useful, especially for complex queries and comparisons. But it still needs to be supported by people doing the original work. Without them, the search machine becomes useless.
Publishers can either bury their heads in the sand and keep optimizing for the blue link. Or they can recognize that it’s endangered, and that the real fight is over keeping a viable habitat for it.