It's not quite like that. The tools used to scrape the web for training data couldn't access the site to scrape the data, so it's not encoded in the model.
The query interface for the model just hallucinates when there's a 'vacuum'.
I was super excited when I realized season three was on Netflix and would pick up off the banger of an ending to season two, got half way through it thinking: this is awesome and weird, it's like season two but slightly different. Then realized I'd already watched season three and was watching it again.
It took me a minute, but now I can't unsee it.