Nieman Lab: How AI is creeping into The New York Times

What the article is about

Nieman Lab senior editor Laura Hazard Owen examines how AI tools are being built and adopted inside The New York Times newsroom. The article centers on the Times’ internal AI tool called “Cheatsheet,” which was rolled out to every journalist in the newsroom by February 2026, and on the broader question of how the organization is managing AI detection in a moment when such detection is technically unreliable.

Context: the Cheatsheet tool

Cheatsheet is an internal AI system built by the Times’ AI Initiatives Team. It is not a general-purpose chatbot but a purpose-built tool for newsroom research tasks. Reporters have used it to investigate an election-interference group, to transcribe and translate Syrian prison records, to analyze Trump statements about January 6, and to compile a cross-media analysis tracking Dr. Mehmet Oz across 2,500 appearances. The same team also built a separate tool — the “Manosphere Report” — to help reporters track online radicalization communities.

These are not writing tools in the conventional sense. They are research and analysis tools that help reporters do more investigative work with structured or multilingual source material than would otherwise be feasible. The Times appears to be drawing a deliberate line between AI as a research accelerator and AI as a writing substitute.

Key argument on AI detection

Owen’s article also covers the Times’ approach to detecting AI-generated content in submissions. Pangram Labs CEO Max Spero acknowledged that AI detection tools both miss AI-generated content and generate false positives, and that the percentage of AI material in a text is difficult to estimate. The models that generate content and the tools that detect it are evolving at the same time, making detection an ongoing arms race rather than a solved problem.

Who should read this

Journalists, editors, and writers at news organizations thinking about how to structure internal AI tools for reporting tasks rather than writing tasks — and anyone tracking the state of AI detection in newsrooms. The article offers a concrete, institution-specific picture of how a major legacy newsroom is making distinctions between AI use cases it accepts and those it does not.