Okay, but when those sites go out of business, where does the AI get its information from? This is obviously not sustainable.

It's only the advertising-funded sites that go out of business and a lot of those sites were in any case just scraping other sites. What proportion of reliable online information is only available from a web site that is funded by advertising? It's not zero, but it's not a very big number, either, I suspect, so it might be sustainable.