This is a useful take: I too will use LLMs for search-- but not for search for journal articles with data and evidence. LLMs too easily confabulate these.
LLM-as-search is fantastic when you want a no-bullshit statistical result for what you’re looking for when you’re wanting an overview or interactive tutorial.
As long as it has footnoting so I can see where each piece of information was sourced from, AI chat has its use cases. Without that I genuinely do not see the point at all. It’s like when people “ask Google” something and just blindly trust the highlighted “answer” as infallible truth. It’s just a really, really bad habit to develop and I wish more people understood this.
Not infallible truth. But very often it’s something that is just for personal use.
Some things I’ve asked it recently were like “Which torch is smaller out of these 5 models?”. Once I find which one I want it’s easy to verify. Or “what does this Spanish expression mean?” or “how do I do …”.
Not everyone uses it to try and write authoritative stuff. And Google is full of clickbaity “comparison sites” that are nothing but fake advertising.
All of those questions you asked it return authoritative answers which you take on face value, unless you spend extra time fact checking them yourself.
