LLM's are just summarizing search results and not always well. This is like asking if academic papers are relevant when college essays exists, imo. I would much rather read the original source of the material than the summarized (potentially badly) LLM result.

They are relevant to me, because I often look for documents rather than answers.

I think the market segmentation is:

- Seeking documents: search engine

- Seeking answers: LLM

Indeed, Perplexity is so much more useful to me, because I usually have a question, that I can just enter and get an answer that usually is correct. And it also gives sources so I can check the answer and find more info. Try e.g. typing "what does defer do in Zig" into a search engine, one of the last questions I asked. Perplexity immediately gave a clear answer.

I only use a search engine to find a specific website if I know the website should exist but don't know its URL, e.g. the website of a festival.

Yes, for me.

Work: “foo corp privacy policy”, “foo okta integration”, “foo corp trust center”

Here I want specific links to exact documents because the details matter.

Home: “3d print boston terrier”, “cyberpunk fight chimera”, “!w georgia country”

Here I want to look at lots of similar things quickly, or read long-form content.

And with Kagi, “cyberpunk fight chimera?” gives me a consistently decent AI summary of the search results, with references. It’s a single UI that gives me exact results or broad summaries as requested, and that’s pretty handy.

Yes, because LLMs make stuff up all the time. Still. Anybody who's using an LLM for search isn't serious about the answers.

Kagi do have an LLM offering if that's what you still want. While still benefiting for your own search results curation and being more privacy orientated. I'm very happy that I can basically have the best of both worlds from them.

LLMs heavily rely on traditional search engines behind the scenes.

Almost all LLM do traditional web search in the background anyway. So if you want data straight from the source without probabilistic text generation, it will be faster to do the search yourself. Like if I want to know the country of origin of a company, would type its name and go to Wikipedia. While LLM would usually search literally "What is the country of origin of ..."? and then visit LLM generated sites that SEO optimized the query on which a fraction of text is going to be relevant after which the information is very low quality.

With Kagi, anyway, search engines are very relevant to me. For my use cases, LLMs provide far inferior results in exchange for requiring more time and effort from me.

FYI, Kagi has Assistant, the chat interface that do the search and summary for you, and you can custom it, limit the search scope.

Yes, because I want actual answers, not something the LLM synthesizes which may or may not have any relation to the truth. LLMs suck for finding info if you care about the accuracy of the result (which I do).