Will AI Overviews reduce search referral traffic to publishers?

The Reuters Institute's 2026 trends and predictions report surveys editors, CEOs, and digital executives at news organizations and finds that publishers "expect traffic from search engines to decline by more than 40% over the next three years," a figure restated elsewhere in the report as traffic set "to almost halve (-43%)." Expectations vary widely around that average: "a few optimists" put the coming impact below 20%, while "around a fifth of respondents expect a loss of more than 75% of their company's search traffic." The forecast sits alongside a broader confidence drop — only 38% of respondents feel confident about journalism's prospects generally, "22pp lower than four years ago" — even though confidence in their own specific business holds closer to steady, at 53%.

The mechanism the report names is Google's AI Overviews and the separate "AI Mode" tab, now live in 120 markets with Gemini-3 multimodal capability. AI Overviews already appear atop roughly 10% of US search results, and the report cites studies showing "zero-click searches" — where the user never clicks through — rise sharply once an overview is shown. The excerpt is careful to size the chatbot side of this separately: using Chartbeat data, it notes ChatGPT referrals are rising quickly in relative terms but remain "little more than a rounding error" in absolute terms, since "Google delivers 500 times as many referrals as ChatGPT from search alone; 1,300 times as many" counting Google Discover. So the decline publishers forecast is specifically about search-answer substitution inside Google's own results, not a wholesale migration of referral traffic toward chatbots.

This gives a concrete real-world channel for the open question in Why did AI article share stop growing after 2025?: if AI Overviews are increasingly answering queries in place rather than linking out, reduced click-through could suppress visibility for AI-labeled and human-labeled articles alike, which would read as "weak search performance" without distinguishing content origin — a confound this excerpt doesn't resolve either. It also supplies the publisher-economics counterpart to How much of the internet is AI-generated now?: where that note measures the web's content filling with AI output, this source names what's draining from the other side of that same shift — search referral revenue to the publishers whose work increasingly feeds those answers.

What the excerpt doesn't establish is the status of its own central number. The 40-43% figures are self-reported expectations from a sample of news executives whose total size and sampling frame beyond "editors, CEOs, and digital executives" the excerpt doesn't disclose — a forecast of sentiment, not a measured outcome. That differs in kind from the one measured data point offered, Chartbeat's finding that "aggregate traffic to hundreds of news sites from Google search has already started to dip," which is real but left unquantified here. The implication the evidence actually supports is narrower than the headline number: search-referral erosion for publishers, especially those dependent on lifestyle content, is already underway and the zero-click mechanism is well evidenced, but the specific 40-43% magnitude rests on executive sentiment about the next three years rather than on extrapolated measurement.