I believe usually when someone complains about text written by a language model they are hoping to read human-written text instead of human-laundered LLM output.