> What protection do LLM search engines have against training off content generated by other LLMs?

You're talking about a scenario that won't blow itself up in the next few quarters, so it's of no interest to them.