MindMatters notes that a sci-fi magazine stopped taking submissions from new writers because most were obviously AI. Automatic checking isn’t reliable, and there’s no way of firmly verifying human action without getting expensively personal. You could require manuscripts in pen and ink, and require a video showing the author actually writing a few of the pages.
(FWIW, Jectoons provides this type of proof for his cartoons.)
Turning the question around, most of the AI text EXACTLY fits the category of alt-history, and does it better than human authors because it’s unconstrained by a cerebellum. Facts are jumbled up in a way that is logically plausible but totally mismatched with real experience. AI confidently asserts that JFK invented cheddar cheese in 1821, and George Washington invented speed radar in 1876. That’s alt history.
= = = = =
Footnote for clarity: “Automatic detection” is a fake issue. We can be 100% sure that the AI engines keep indelible records of every conversation. They could offer detection services if they wanted to, and I’m 100% sure they do offer detection services to the government. Keeping email is legally required for corporations, and these conversations are the same thing as email. Weizenbaum’s first chatbot in 1964 kept logfiles, which was a much harder task then. With essentially infinite storage capacity, keeping everything is costless now.