They tell students papers sound like AI — even when they're the students' own writing — so the company can scare them and sell them a product.
This goes beyond the general problem that even good AI detectors aren't perfect and shouldn't be the sole reason to accuse someone of writing with AI.