The Associated Press AI rules remember that fluency is not journalism

There is something refreshing about a corporate policy that refuses to be dazzled. While software announcements present every AI feature as a civilizational hinge, the Associated Press has managed to draw a much cooler, brighter line: generative AI output is unvetted source material.

In its standards around generative AI, the AP states that while staff can experiment cautiously, they may not use AI to create publishable content. Furthermore, the news organization strictly forbids using generative tools to add or subtract elements from photos, video, or audio.

This isn't technophobia; it’s an anti-confusion mandate. A journalist's byline is a public promise that the work has passed through a gauntlet of reporting, verification, editing, and institutional accountability. Generative AI strains this promise because it can produce highly fluent text without doing any of the underlying civic labor.

Fluency is not the scarce good in journalism; verification is. By placing AI in the correct hierarchy—treating it like a tip sheet, a prompt, or a tool to organize known facts—the AP avoids letting the software impersonate the work. It recognizes that newsrooms have always used technology, from telegraphs to spreadsheets, but the machine doesn't get to graduate into an author.

A label warning that 'AI helped write this' means very little if the publication cannot explain who verified the facts. Journalism doesn't begin when the sentences sound authoritative; it begins when a human takes responsibility for knowing whether those sentences are actually true.