Google's AI Overviews rolled out to U.S. search users in May 2024, one year ago this month. The Reuters Institute's January 2026 Journalism, Media, and Technology Trends and Predictions report, surveying 280 digital news leaders across 51 countries, found Google search referrals to news sites fell 33% globally between November 2024 and November 2025 — and 38% in the United States across the 797 U.S. sites tracked by Chartbeat. [1] Publishers in the survey expect search traffic to fall 43% over the next three years; about one-fifth expect declines above 75%. [1]

Digital Content Next, the U.S. trade association whose members include the New York Times, the Washington Post, NBC News, and the Wall Street Journal, has put the same period at a 10% median year-over-year decline in Google referrals to premium publishers, with member-level reports ranging from 1% to 25% by category exposure. [2] DCN CEO Jason Kint posted the number publicly: "Since Google rolled out AI Overviews in your search results, median year-over-year referral traffic from Google Search to premium publishers down 10%." [2] The gap between the Reuters Institute's 38% U.S. number and DCN's 10% premium-publisher median is the category-exposure gap: lifestyle, weather, recipes, TV guides, and how-to publishers have absorbed steeper losses than hard-news brands, because Google has partially exempted hard-news queries from AI summaries. [1]

The platform's defense is the per-click quality argument. Google's developer documentation says: "We've seen that when people click from search results pages with AI Overviews, these clicks are higher quality (meaning, users are more likely to spend more time on the site)." [2] The defense is separable from the volume argument. A 25%-to-38% drop in clicks paired with higher dwell time is consistent with a smaller, better-converting click stream. It is also consistent with an ad-revenue collapse on pages whose business model is impressions, not conversions. Kint reads the gap between Google's stability framing and the dashboards as a measurement dispute, not a definition dispute. [2]

This story belongs in culture because the Wednesday read is not about Google's revenue. When the dominant channel through which strangers find a publisher's work shrinks by a third or more in a year, the publisher does not lose a third of its readers — it loses a third of its discovery surface. The first-time reader, the long-tail evergreen, and the recipe-utility traffic are differentially exposed. The signed-in subscriber and the newsletter loyalist are not.

The publishers' strategic pivot is underway. The Reuters Institute survey found news organizations planning to increase investment in original investigations (+91%) and contextual analysis (+82%), and to cut general news (−38%). [4] Seventy-nine percent planned to invest more in video, 71% in audio. [4] "Journalism's best response is to double down on the things that make us valuable and unique," said Taneth Evans, head of digital at the Wall Street Journal. [4] The consequence is a labor-force question — the journalism the AI summary cannot replicate is more expensive per byte than the journalism it can.

Inside the chatbot ecosystem, the publisher is now a citation. ChatGPT delivered 0.02% of all publisher referrals as of November 2025, versus Google Search's 7.3% — a 365-to-1 gap. [1] Google delivered roughly 500 times the search referrals as ChatGPT, or 1,300 times once Google Discover is included. [1] The shift from search to answer-engine has not produced a comparable referral channel.

The fee-licensing model — OpenAI's 2025 and 2026 agreements with the Guardian, Schibsted, and Axios; Amazon's deals with the New York Times, Condé Nast, and Hearst for training rights — is the publishers' settlement instrument. [4] The institutional question is whether an industry whose discovery surface contracts 25% to 38% year-on-year can finance journalism out of licensing fees, or whether the licensing-fee era is the interim before another instrument.

The one-year mark is the right date to print. Year two is the year the licensing settlements get tested. Year two is also when DCN's 10%-to-25% turns into 20%-to-50% or stabilizes inside its current band, and when the discovery surface either consolidates around fewer publishers or fragments across newsletters, podcasts, and direct apps. The Reuters Institute's chart is the platform's measurement of its own externality. The question is whether the publishers print it next year as a denominator or as a postscript. [3]

-- ANNA WEBER, Berlin