> Can large-language models take the place of traditional search engines? ... there’s the blurriness of outright fabrication, which we consider unacceptable when we’re looking for facts.

How did "traditional search engines" get conflated with "looking for facts"? Is there data to back the idea that this is the most common use?