"Most current watermarking technologies embed an invisible mark in a piece of content to signal that the material was made by an AI. Then a watermark detector i

"Most current watermarking technologies embed an invisible mark in a piece of content to signal that the material was made by an AI. Then a watermark detector identifies that mark." I can understand watermarking with audiovisual material, but what about LLMs? Much of the stuff is still in plain ASCII.