I suspect it because search results require manually parsing through them for what you are looking for, with the added headwinds of widespread, and in many ways intentional degradation of conventional search.
Searching with an LLM AI is thought-terminating and therefore effortless. You ask it a question and it authoritatively states a verbose answer. People like it better because it is easier, but have no ability to evaluate if it is any better in that context.
BTW, all the modern LLMs I’ve tried that do web searching provide citations for the summaries they generate. You can indeed evaluate the validity of their responses.
I suspect it because search results require manually parsing through them for what you are looking for, with the added headwinds of widespread, and in many ways intentional degradation of conventional search.
Searching with an LLM AI is thought-terminating and therefore effortless. You ask it a question and it authoritatively states a verbose answer. People like it better because it is easier, but have no ability to evaluate if it is any better in that context.
So it has advantages, then.
BTW, all the modern LLMs I’ve tried that do web searching provide citations for the summaries they generate. You can indeed evaluate the validity of their responses.