>By scraping TikTok and Discords?

One of two ways. Yes, by scraping, even it it requires users to 'sell' their own browsing data to the AI companies because places like Discord lock them out.

Or, the other way is for particular event organizers to pay directly for their services to be advertised/incorporated into the LLM itself. Those that don't pay get more and more of their data erased from the LLM maybe?