Just a note regarding LLM (ChatGPT). Do not treat them as search engines (search engine is big database where you remember the source of your information plus the exact content). Consider it as a big memory (closer to human memory). It learns patterns, most common sequences etc. It has no idea where the information came from. So you cannot implement your idea with actual LLM. You will endup with classical search engine.

Often the LLM blends the information, mixes it and even halucinates nonsense. Examples where LLM answers the exact "copyrighted" text is probably rare and it can happen seeing the text many times. I guess it can do good replicas of Shakespeare, but not of your local writter.

To understand the LLM more.. just remember that is does not have any logic. It is just memory. It does not understand math and coding. It remembers 1+1=2.

Your idea has some interesting points. How to treat the middle man. Something learns from your content (not copies it) and then sells it. And my question is.. Can I trust this middle man?