the context of the modern LLM's are large.. and what people do is they augment it with their own data so it's not as generative and working off their sources. Text summation is literally one of the fundamental uses that came from the language models.