Raw LLM Responses

That's way overstating your argument. Ai is pretty good at infilling spaces between known cases. eg. if you have case "1" and case "3", then AI can infill case "2" between them, which might never have been done by a human before, so is new content. It can extrapolate new cases in a series too, so given  situations we might summarise as 1, 2, and 3 it could give you 4. It probably won't (yet) give you completely new, non-derivative material, but it's still in its infancy, so who knows what's imminent. And how much human content is derivative anyway. Many people aren't so different !