> I focus the message box to draft a reply, but there’s already one there. It was also generated by the language model. I delete it, replacing it with my own.

The really, really scary thing is how uncommon this approach is. I think.

My assumption is that most people roll with automated pre-written reply. Maybe tweaking a few things here and there, but ultimately preferring the all-too-convenient trade-off of the robots having written something close enough to what they wanted to say, using "better" words. Even when what they would have written themselves would have had some personality, even if it was their own flawed human one.

For the record, I am 100% with you on your approach (on the odd occasion that I must use gmail).