There’s a lot of other topics about this already (see 1, 2, 3, 4, 5, and more in the “AI” tag), so I don’t want to hijack this one. Happy to discuss more in any of the other topics, though!
While I don’t have the exact stats to prove it (I don’t know what % of Kagi’s results come from their own index vs Marginalia’s, for example), it feels sufficiently different to me to qualify as its own search engine instead of “just another aggregator”.
And their customization options and lack of ads make them pretty unique, search sources notwithstanding. It’s a degree of personalization that no other engine, as far as I know, offers. I block a few dozen spam domains, deprioritize a few others, and prioritize a few usually-good sources, and that makes a pretty big difference in result quality (though you can accomplish similar with Google and uBlacklist).
Kagi can then use those raise/lower/block signals in aggregate to refine their default results. (Seems like its users really, really hate Pinterest and strongly dislike junk news sites.)