You see the same with code and text generated by LLMs. Overly verbose, comments that repeat code, and commit messages that repeat what is done in code, but not WHY.