i know that this is probably a very bad approach, but until now it worked out - if a text contains grammatical errors i automatically assume that it was written by a human, as AI-content always sounds so polished, boilerplate-y and corporate. but you can probably tell your AI to integrate errors, ignore capitalization and miss out on some punctuations.

I asked ChatGPT to reply to your comment and to include some errors, leave out some punctuation etc., that's the result:

oh, i see what you're saying. yeah, it's really hard to tell if something is bot-written these days. like, even reading your comment, i start to question if i'm talking to a human or not. the examples you mentioned do stand out, but more because they're so similar, not necessarily bot-like. and you're right, the "firehose of misinformation" is just getting worse and worse. it's a strange new world we're living in.

i think we're cooked.