The problem runs even deeper than you describe.

A few days ago I ran across a news piece about a guy who was in a Zoom call with his boss and several coworkers.

The boss directed him to carry out a financial transaction. He had a transitory sense that something seemed off, but everyone else on the call took it in stride, so he did as he was told.

As it turns out, he was the only human on the call. Everyone else was a deep-fake video based on AI. The transaction he completed gave the people behind the scam a huge amount of money (I can't recall if it was tens of millions of dollars or more).

So, I'm far less concerned these days about detecting what piece of writing was drafted by AI or even written and completed by AI. The bad actors are getting so sophisticated using AI that we have to figure out secret signals to let our colleagues, friends, and even family know when it's really us.