Yeah, but in the long run (or very soon), you might have to use grammar misstakes as a captcha, to seperate bots from humans.

> misstakes > seperate

I see what you did there.

Please never implement that idea though, it would be like the opposite of https://xkcd.com/810.

I am pretty sure, it is already in use.

But of course it would be trivial to modify the llm output, to include random misstakes, if this ever would become a standard way of dealing with bots.

> Spam filter idea: posts with bad interpunction are likely content-free angry ramblings!

... or written by non-native speakers.