Back in June I wrote in this newsletter about not using an AI detector on First Opinion submissions.

How things have changed in just three months.

This week, I officially began running First Opinion submissions through Pangram, which I’m told is the most reliable AI checker commercially available right now. I don’t feel great about it: I worry about false positives, particularly on pieces written by those for whom English is a second language. And it still feels a little hypocritical. But frankly, I needed to do something.

The results have been illuminating. I am not automatically rejecting every piece Pangram flags. But it helps me start the “Did you use AI on this?” conversation, which is always an awkward one. (Interestingly, I’ve found that sometimes authors are using AI without disclosing it to a communications person who submits a piece to me — and vice versa.)

When people complain about our policy of not permitting any AI-generated text, I always want to say: Just three years ago, you would have written this piece without Claude, Gemini, or ChatGPT. You can still do it! Don’t let your ability to write atrophy.

Recommendation of the week: The New Yorker tells the story of a dressage coach who was found not guilty by reason of insanity for the attempted murder of his most aggravating student. If courts went by the framework of Reddit’s “Am I the A-hole?” sub (as they should, of course), the verdict would have been “everyone sucks here.”