Google’s AI answer engine is quietly strangling the open web

Google’s new “Preferred Sources” button is presented as a modest tool for publishers, but the article treats it as an admission that the company’s relationship with the open web has broken down. The feature lets readers flag favorite publishers across Search, Discover and Google News. Its arrival, the article argues, comes as Google increasingly answers questions itself instead of directing users to the websites that produced the information.

For 25 years, the basic bargain was straightforward: publishers allowed Google to crawl and index their work, and Google sent them readers. Those readers supported advertising, subscriptions, journalism, reference works and small businesses built around organic search. The article says that bargain is now unravelling in real time.

Chartbeat data from more than 2,500 news sites worldwide showed Google search referrals falling by about one-third in the year to late 2025. American publishers suffered an even larger decline of about 38%, according to the figures cited. SparkToro estimates that zero-click searches now make up roughly 68% of all Google searches in 2026, compared with 60% in 2024 and 49% in 2019.

The article cites a Seer Interactive study covering thousands of search terms through late 2025. It found organic and paid click-through rates had fallen 61% and 68%, respectively, since AI Overviews became widespread. Even cited websites averaged an organic click-through rate of only 0.70%, down 49.4% year over year, while uncited sites averaged 0.52%. Searches without an AI Overview produced a traditional rate of 1.45%.

The damage has reached individual companies. HubSpot estimated that it lost 70% to 80% of its organic search traffic, while Chegg reported a 49% annual decline in non-subscriber visits. DMG Media told the U.K. Competition and Markets Authority that click-through rates fell as much as 89% on desktop and 87% on mobile when an AI Overview appeared above its link. The article frames the shift as a structural rupture, not a routine algorithm update, and begins to turn toward how smaller and midsized outlets might respond.

“

extinction-level event