No, a lot of them have LLMs integrated into their search engine, but they themselves are not LLMs.
The search engine I use (SearXNG) is what’s called a meta-search-engine, meaning it runs my query through several different search engines (which I can toggle on or off at will), then parses the results from all of the queries engines, strips out all of the slop bullshit, and returns me actually relevant links.
I think if you wanted to you could attach an LLM to a self-hosted SearXNG instance, but I’m not putting LLM fuckware on my system. I prefer my computations to be deterministic.