Their philosophy is that false positives are much worse than false negatives (where "positive" means AI generated), so one would expect high precision and only medium recall.

I had the same experience for my writing that was AI generated.