There’s something funny going on with AI and op-ed writing.

It’s not just that people are using it. At this point, I expect that. It’s that not everyone involved with an op-ed knows it’s happening.

I work with a lot of First Opinion submissions that involve several people. There might be two or three authors on it, or perhaps it comes to me via a PR company or the communications department at a university. When I run the piece through Pangram, the AI detector, it lights up as AI-written. Then I tell the person I’m corresponding with, and they’re shocked. They deny any AI use at all.

But a few days later, I find out: Yes, someone involved did use it. Perhaps one co-author didn’t tell the others, or the author didn’t tell the university comms person who submitted it to me.

So: Have those hard AI conversations before submitting a piece to me (or any other opinion editor)!

Relatedly: I’m also seeing people who use AI sort of without realizing they’re really using AI? They think it’s just “proofreading” or “smoothing out sentences.” AI has a pretty heavy hand, though, and often ends up rewriting quite a bit — and authors don’t always look at the output closely enough to see how much the language has changed.

Recommendation of the week: In New York magazine, Rachel Corbett speaks to Mark Fowler, the physician connected to the failed execution of Christa Pike in Tennessee. Fowler has been involved with several executions, which he does not see as a violation of the Hippocratic Oath. Corbett writes: “For him, the work is about the fulfillment of a solemn obligation. ‘How can you morally say, “I’m for the death penalty so long as somebody else does the work?”’ he says. ‘I do things that nobody wants to do.’”