LLMs are still notorious for hallucination; last I checked ChatGPT in particular still hallucinates about 1/3rd of the time.
So yeah, I do still use search engines, specifically Kagi and (as a fallback) DuckDuckGo. From either of them I might tack on a !g if I'm dissatisfied with the results, but it's pretty rare for Google's results to be any better.
When I do use an LLM, it's specifically for churning through some unstructured text for specific answers about it, with the understanding that I'll want to verify those answers myself. An LLM's great for taking queries like "What parts of this document talk about $FOO?" and spitting out a list of excerpts that discuss $FOO that I can then go back and spot-check myself for accuracy.
So yeah, I do still use search engines, specifically Kagi and (as a fallback) DuckDuckGo. From either of them I might tack on a !g if I'm dissatisfied with the results, but it's pretty rare for Google's results to be any better.
When I do use an LLM, it's specifically for churning through some unstructured text for specific answers about it, with the understanding that I'll want to verify those answers myself. An LLM's great for taking queries like "What parts of this document talk about $FOO?" and spitting out a list of excerpts that discuss $FOO that I can then go back and spot-check myself for accuracy.