Skip to content
NewsIncident

Microsoft Data Shows Copilot Cut New York Times Referral Clicks by 87 to 93 Percent, Unsealed Court Documents Confirm

· by Pondero Newsdesk

The short version

A motion unsealed September 17, 2026 in NYT v. Microsoft puts Microsoft's own internal analysis into the public court record, showing Bing Chat sent 87 to 93 percent fewer referrals to New York Times sites than traditional Bing searches, with executives privately calling the practice theft.

Microsoft Data Shows Copilot Cut New York Times Referral Clicks by 87 to 93 Percent, Unsealed Court Documents Confirm

A court motion unsealed September 17, 2026 in the New York Times v. Microsoft copyright case puts a specific number on the referral-traffic collapse that publishers have attributed to AI answer engines: Bing Chat sent 87 to 93 percent fewer referrals to New York Times websites than traditional Bing searches for equivalent queries. The figure comes from Microsoft's own internal analysis, not the plaintiffs, per The Wrap's review of the filing.

What the filings contain

The unsealed motion was filed by a coalition of news publishers as part of their copyright infringement case against OpenAI and Microsoft. It draws from internal documents produced in discovery.

Click-through declines varied by publisher. Bing Chat directed 87 to 93 percent fewer clicks to New York Times sites compared with traditional Bing searches. For Daily News plaintiff sites, the drop ran 83 to 91 percent. Across the broader set of publishers covered in the motion, click-through declines ranged from 51 to 94 percent, per Superpower Daily's reporting on the filing.

The documents also surface executive language that complicates each company's public positioning. Brent Hecht, Microsoft's Director of Applied Science, wrote in a January 2023 internal memo that web scraping for AI training amounted to "the largest theft of labor in human history" and would make a "complete mockery" of fair use, per The Wrap. A separate Microsoft document stated: "It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business."

On the OpenAI side, Nick Turley, head of ChatGPT, wrote internally that commercial AI trained on news content was an "existential threat" to publishers and that the resulting products were "largely substitutive." A separate document shows OpenAI employee Nick Ryder discussed bypassing the New York Times paywall for scraping purposes; co-founder Greg Brockman responded affirmatively, per the same filing.

Microsoft CEO Satya Nadella testified that chatbot conversations substituted for publisher website visits, acknowledged paywalled content should have been licensed before AI training, and said he would have required model retraining had he known OpenAI scraped paywalled material.

Why it matters

The significance for publishers is that internal defendant analysis now sits in the public court record at specific percentage ranges. Abstract arguments about AI substituting for referral clicks become harder to contest when Microsoft's own numbers produced the 87 to 93 percent figure.

For anyone running an AI-first content operation, the filing spells out the mechanism plainly: generative AI answer interfaces provide enough of a response that users do not click through to the original source. That substitution effect is the core tension between AI answer engines and the web's referral economy. Any content operator whose traffic model depends on organic referrals from AI-augmented search should treat these percentages as a concrete data point, not a hypothetical risk.

Several large news organizations remain in active licensing negotiations with OpenAI and Google as of September 2026. Those talks now proceed against a court record that includes each side's own internal assessments of the traffic damage. Publishers negotiating deals can cite defendant-sourced numbers to anchor their valuations.

What to watch next

No trial date has been set as of the September 17 unsealing. Watch whether the internal documents accelerate licensing deal closures (by handing publishers credible anchor numbers) or harden positions and push the case toward a trial ruling on the copyright claims.

Sources