AP bans journalists from using ChatGPT to write stories

The gold standard of journalism style just drew a hard line on AI.

Associated Press, whose Stylebook sits on the desk of practically every newsroom in America, just issued formal generative AI guidelines for its staff. The bottom line: reporters can play around with ChatGPT, but they cannot use it to generate publishable stories.

AP vice president for standards and inclusion Amanda Barrett made the stance crystal clear, writing that the news agency does not see AI as a replacement for journalists in any way. The new rules are designed to keep AI strictly in the role of an assistant rather than an author.

Treat AI like an unvetted source

Under the new guidelines, AP journalists are instructed to treat any text coming out of a generative AI tool as completely unvetted source material. That means everything AI spits out must go through the same rigorous, manual fact-checking process as an anonymous tip from a stranger.

Visual and audio standards are even tighter. Staff are strictly barred from using AI to alter photos, videos, or audio tracks. Fully generated AI images are banned entirely, with one narrow exception: when the AI image itself is the subject of the news story. Even then, the caption must explicitly state that the image was synthesized by AI.

Reporters are also prohibited from inputting confidential reporting materials into public AI platforms, and they are instructed to actively verify that outside sources are not feeding them AI-generated misinformation.

The great newsroom paradox

What makes AP's stance so fascinating is the contrast between its editorial rules and its corporate moves.

Right around the time it clamped down on internal newsroom use, AP signed a licensing deal with OpenAI, handing over its vast archive of news stories to train models. At the same time, AP has spent years using basic automation algorithms to churn out corporate earnings briefs and minor league sports roundups.

It represents a completely different playbook from peers like The New York Times, which recently updated its terms of service to block AI crawlers and walked away from coalition talks with AI firms altogether.

Why it matters

When AP speaks on editorial standards, the rest of the industry takes notes. Because so many publications base their internal house styles on the AP Stylebook, these guidelines will likely become the default playbook for hundreds of newsrooms wrestling with how to handle ChatGPT.

For everyday writers and knowledge workers, AP's framework is a practical lesson. Generative AI is fantastic for brainstorming and summarizing, but treating its output as final draft material without rigorous auditing is professional suicide.

We think AP is threading a very delicate needle here. Editorial integrity is the only moat a legacy wire service has left, so treating AI as an unreliable narrator is smart. At the same time, cashing in on its back catalog by selling training data to OpenAI shows AP is not naive about where the industry's money is headed. It is a pragmatic, if slightly hypocritical, survival strategy.