Deep research is still search behind the scenes. The quality of the LLM’s response entirely depend on what’s returned. And I still don’t trust LLMs enough to tell fluff from truth.
Yeah but Deep Research, at least in the beginning (I feel like it's been nerfed several times) would search often on the orders of 50+ websites for a single query, and often times reading the whole website better than what an average human could.
Deep Research is quietly the coolest product to come out of the whole GenAI gold rush.
The google version of Deep Research still searches 50+ websites, but I find it's quality far inferior to that of OpenAI's version.
I do check the RAG sources from deep research, but you're very right in that it's easy to start taking mental shortcuts and end up over relying on LLMs to do the research/thinking for you.
