[dbaman]

But is that first point really true?

If a document is used to train an LLM, that means that an AI checker would be more likely to give a false positive that the document is written by AI?

I guess it depends on the AI checker.

I would think that that LLM, after generating the fragment "When in the course of.." would just be a bit more likely to generate "human" as the next word, yes?