What specifically do you find to be mediocre? I feel like LLMs write better than most people I know, myself included.

There could be a mismatch on what the state of the art really is these days. In my experience, since the release of GPT-4 and especially 4o, ChatGPT has been able to do the vast majority of concrete things people tell me it can't do.

I suspect most people using LLMs to write things are writing a bunch of boilerplate that nobody is going to read anyway. I could be wrong.