Newsrooms that once debated whether to use AI are now debating where it belongs in the production process, and who signs off when it gets something wrong.
At several regional publishers, AI tools now transcribe interviews, suggest headlines and check copy against style guides. The common thread among those doing it well, editors said, is that every AI step sits inside a workflow with a human checkpoint after it.
“The model doesn’t get to publish. It gets to flag,” said Rachel Kim, a managing editor who helped write her newsroom’s AI policy. “A person decides.”
The model doesn’t get to publish. It gets to flag.
Rachel Kim, managing editor
Several editors said they treat AI checks the same way they treat a copy desk: as a stage a story must clear, with clear criteria for what passes.
