Whats funny about LLMs is they are trained from books, but they are also trained to not output books, so how much of an LLM skews its output because a perfectly

Whats funny about LLMs is they are trained from books, but they are also trained to not output books, so how much of an LLM skews its output because a perfectly normal sentence could be a quote in like 300 different books?