English is not my first language. Sometimes, especially when I try to write a complex and long idea, I use LLMs to correct my errors, syntax, etc. In those cases, even if the idea remains the same, the text ends up looking LLM generated, since the correction applies its own "robotic style." This is even more noticeable if you ask chatgpt to translate it from another language.
(I'm not the author of the post, just mentioning it since it will probably become harder to detect such cases.)