September 19, 2026
The web is losing traffic to AI: Copilot cut NYT click-through rates by 83–93%
On September 17, 2026, internal company documents were partially unsealed in The New York Times v. OpenAI and Microsoft. A Microsoft presentation from January 2024 describes a “doom loop” that degrades both models and the web. According to Microsoft, Copilot reduced click-through rates for The New York Times and Daily News articles by 83–93%, and for Ziff Davis domains by 51–94%.
Web scraping once looked like raw material for training to companies: in January 2023, Microsoft director Brent Hecht called it “the largest theft of labor in human history,” and Project Taxi passed OpenAI a compilation of “billions” of Bing pages. Now the documents show the other side: OpenAI's mid-training datasets contained more than 91 692 copies of works from The New York Times, Daily News, and the Center for Investigative Reporting, while Project Mango contained copies of at least 160 903 works by news publishers. Nick Turley wrote that OpenAI products are “largely substitutive” and will become even more substitutive as they improve.
The market faces a simple conflict: models replace the sources from which they get data for the next generation.
