If AI-generated code is considered acceptable in your project then you aren't using a powerful-enough programming language. And you're paying the cost in code bloat.
How many Coq programmers find ChatGPT useful? How many nontrivial Coq programs written (and not merely memorized) by ChatGPT even pass type checking?
If you're considering AI-written code then you have a bloaty-code problem. Letting an AI write bloaty code conceals the symptoms (keystroke count) but tech debt will still kill your project sooner or later.