For publishers, it has long seemed obvious that when AI answers a user’s question directly, there is less reason to click through to the source. Over the past year, research has repeatedly confirmed it: controlled experiments with search users, analyses of millions of search queries, studies of click-through rates, and traffic data from thousands of publisher sites all point in the same direction.

The methodologies and measurements differ. But the evidence is clear: When AI-generated answers appear in search, fewer people click through to the sources behind those answers. And while referrals from standalone AI platforms are rising, they remain far too small to offset the traffic publishers are losing from traditional search.

That conclusion stands in contrast to claims that AI-powered search can create more opportunities for sites to be discovered and send publishers higher-quality traffic. As DCN VP of Research Rande Price recently wrote, experimental evidence finds little support for those claims.

No single study can capture the full impact of a discovery ecosystem changing this quickly. But with researchers approaching the question using different methodologies, datasets and measures, declining referral traffic can no longer be dismissed as anecdotal, or AI’s role in that decline as speculation.

AI answers are reducing clicks

Some of the clearest evidence comes from research designed to isolate the effect of AI answers from everything else happening in search.

In a randomized field experiment, Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon University studied the browsing behavior of more than 1,000 Google users. When an AI Overview appeared, outbound organic clicks fell 39.8% and zero-click searches increased 34.5%.

Research based on observed search behavior reaches much the same conclusion. Pew Research Center analyzed nearly 69,000 Google searches made by 900 U.S. adults. Users clicked a traditional search result on 8% of visits when an AI summary appeared, compared with 15% when one did not. Links within the AI summaries themselves attracted clicks on just 1% of visits. Pew Research Center

And the pattern holds at much larger scale. Ahrefs analyzed 300,000 keywords and estimated that the presence of an AI Overview reduced click-through to the top-ranking organic result by 58%. The effect extended down the results page, with estimated declines of 51% for the second-ranked result and 46% for the third.

It is important to note that these studies differ: a controlled experiment, observed browsing behavior, and search click-through data. However, it is clear that each reaches the same conclusion: When AI answers appear, publishers get fewer clicks.

AI referral traffic is growing, but its not nearly enough

The traffic flowing from AI platforms is growing quickly. It is also nowhere close to replacing the search traffic publishers are losing.

Reuters Institute’s Journalism and Technology Trends and Predictions 2026 puts the gap in stark terms. Drawing on Chartbeat data from 2,576 publisher sites, the report found that Google Search referrals fell 33% globally between November 2024 and November 2025, including a 38% decline in the U.S. However, the researchers caution that AI Overviews cannot account for all of that decline; algorithm changes, publisher strategies and shifts in audience behavior also play a role.

But the scale of AI referrals offers little evidence of a replacement channel. In the same dataset, Google Search accounted for 7.3% of publisher referrals and Google Discover another 13%. ChatGPT accounted for just 0.02% and Perplexity 0.002%. Google Search alone delivered roughly 500 times as much referral traffic as ChatGPT.

Sensor Tower’s State of Web 2026 shows how misleading the growth rates can be. Traffic from AI assistants grew 86% year over year in 2025, while generative AI referrals still accounted for only about 0.7% of overall web traffic.

Similarweb’s research on AI search finds the same disconnect from another angle: use of AI platforms continued to climb while outbound referral traffic plateaued. More people using AI does not necessarily translate into more traffic for the sites supplying the information.

Fewer opportunities to monetize and build relationships

The impact of fewer referrals extends well beyond the lost pageview. Every visit gives a publisher the opportunity to serve advertising, convert a subscriber, register a user, promote a newsletter or app, collect first-party data, and build a direct relationship with that person.

AI search changes that exchange. Publisher content can still contribute to satisfying a user’s query, but the visit that historically followed discovery may never happen. The platform retains the user while the publisher loses the opportunity to monetize the interaction or deepen the audience relationship.

That distinction matters because traffic is not simply a distribution metric. For publishers, it is one of the mechanisms through which investment in original content produces economic value. When AI systems satisfy more demand without sending users to the

AI search is breaking the publisher-platform value exchange

The research now spans controlled experiments, observed user behavior, analyses of hundreds of thousands of keywords, and traffic data from thousands of publisher sites. The measurements vary, but the finding does not: AI-generated answers reduce clicks to publishers, while traffic from AI platforms remains far too small to replace what is being lost.

This changes more than the mechanics of search. For decades, search created an exchange: publishers made their content discoverable, platforms helped users find it, and publishers received traffic in return. AI search is breaking that exchange.

AI systems can increasingly use publisher content to answer a user’s question without delivering that user to the publisher. The content still creates value. What is disappearing is the mechanism that returned some of that value to the companies that invested in creating it.