AI Summaries in Search and Their Impact on Publisher Traffic

When Google answers the question for you, do you still click through to the websites that supplied the answer? A field experiment with more than 1,000 users finds that AI-generated summaries cut organic clicks by around 39.8% — without making search any more satisfying.

For most of the history of web, search has run on a basic caveat: search engines help users find information and the websites that carry it, and those websites receive traffic in return. Every click was a visitor, an opportunity to earn advertising revenue, attract subscribers and build an audience.

The arrival of AI-generated summaries, such as Google's AI Overviews, threatens to usurp that bargain. These summaries synthesise content from multiple sources and present a comprehensive answer prominently, often at the very top of the results page. The feature is designed to make information faster to reach. However, it raises an uncomfortable question. If users are satisfied by the summary itself, they may never visit the sites that produced the information. Falling footfall to these sites might question their survival. Who, then, will keep creating the information in the first place?

To find out how this shift changes behaviour, the researchers ran a field experiment with more than 1,000 participants in the United States. A custom browser extension let them observe real-world browsing over two weeks. Not everyone saw the same thing: some participants were shown Google's results with AI summaries, while others saw the same results without them.

The contrast was stark - where AI Overviews appeared, users clicked through to external websites far less often organic clicks fell by roughly 39.8%, and the likelihood of a "zero-click search" (one in which the user never leaves Google) rose by about a third. In practical terms, users were content with the answers on Google's own page and saw no need to visit the original source.

The consequences of such action extends well beyond any single website. Publishers, journalists, educators, bloggers and countless other content creators depend on search traffic; each visitor who stays on Google, satisfied by an AI summary, represents revenue that may simply disappear. The effect was strongest when AI Overviews sat at the top of the page, drawing attention away from the highest-ranked results, precisely the links that traditionally receive the most clicks.

Might AI Overviews at least encourage people to search more? The researchers found no evidence of it. Some proponents argue that easier access to information should boost engagement, yet search volume remained largely unchanged while clicks declined.

On further examination, researchers failed to find evidence of AI Overviews generated “higher quality” visits to publisher websites. Users who clicked through after viewing AI Overviews,

spent similar amounts of time on websites, exhibited comparable bounce rates and were equally engaged than users who searched without AI summaries. Suggesting, the decline in traffic is not compensated by more valuable visits for publishers.

The most surprising finding concerned satisfaction. One might expect that stripping away AI summaries would make search feel less convenient or less useful. In fact, users who saw the AI summaries reported almost identical levels of satisfaction, information quality and ease of finding answers as those using Google's standard interface. Put plainly: AI Overviews reduced traffic to publishers without improving the search experience.

The researchers also examined Google's emerging AI Model - a more conversational, generative search experience. Although these results are preliminary, they point to an even stronger trend: users engaging through conversational AI generated substantially fewer visits to external websites than those conducting traditional searches.

If these patterns hold, the future of search may look very different from the web we know. For years, search engines acted as gateways, steering users out toward the wider internet. Generative AI is gradually turning them into destinations, where users receive their answers without ever leaving the platform.

Supporters argue that this evolution saves time and improves convenience. Critics counter that it concentrates power within a handful of technology companies while eroding the economic foundations of independent content creation. The debate is already drawing regulatory attention in the United States, Europe and the United Kingdom, where policymakers are asking whether platforms should be required to compensate the publishers whose content helps train and power AI-generated answers.

The research does not settle those debates. But it offers some of the clearest evidence yet that AI-generated summaries are reshaping how attention flows across the internet. As generative AI becomes more deeply embedded in digital platforms, society may have to confront a difficult trade-off: the very technologies that make information easier to consume could also undermine the ecosystem that produces it. How that balance is struck may ultimately decide the future of the web