One month with Kagi search
Toward the end of August, I signed up for a trial of Kagi -- a privacy-focused search engine.
You get 50 free searches, and then, if you want to continue, you can convert to a paid account at $10 a month.
I mentioned here that I wasn't planning on converting to paid, as $10/month felt very steep and I didn't think I could make it my default search on my iOS phone, but @pallas's comment here ultimately made me want to give it a try.
Thus, I dropped the $10 bucks to turn the free trial into a paid one-month trial.
I'm very glad that I did.
The free trial itself was actually not very convincing to me. Knowing that I had limited searches and not wanting to run through them more than I needed, my searches were in the single digits each day. I was very judicious about what I searched and how I typed it. Furthermore, I kicked myself if I instinctively typed something like "imdb everything everywhere all at once" into Firefox's search bar instead of going to imdb.com and then typing in the movie title, as that meant I'd wasted 2% of my allotment on what wasn't technically a search but more of an internet navigation optimization.
On the searches I did I felt like I got good results, but I wasn't sure if that was because of the quality of the service or if it was because I'd simply thought more about what I was actually typing in. Also, the trial made me way too aware that I was searching with limited queries to really make me feel at ease about actually using the service.
Now that I've paid for a month, however, I've just used it as a stand-in for how I used to use DuckDuckGo -- "wikipedia steam deck"-style searches and all.
Kagi doesn't track your search contents, but they do track your number of searches. I have completed roughly 400 searches this month, which Kagi says costs roughly $5.00 out of the $10.00 that I paid them. I don't know nearly enough about any of this to know whether this is an accurate accounting of actual costs or overstating things, but I will say that the $10 price that I initially felt was steep has looked a lot more worth spending after a month on the service.
Kagi generally finds what I'm looking for within the first link. If it's not the top link, it's in the top 3. Furthermore, it seems to dredge up less junk. With DuckDuckGo, I loved that I wasn't being tracked for the purposes of advertising, but it felt like DDG had no problem serving me pages that were built specifically for that purpose. I'd often look up product reviews and get re-routed to sites that appeared to be nothing more than machine-generated lists of recommendations with Amazon affiliate links. I've had to deal with less of these while on Kagi. Some of them still come up, but they're either further down the rankings or they're put into their own "Listicles" section.
Where Kagi really shines though, is local searches. Pretty much the only time I would bang through to Google from DDG was for local stuff. I don't know if it's my location in particular, but DDG is not great about giving me things that are specific to my area, often preferring to give me a smattering of things that are from similarly named locales from elsewhere around the world. Kagi, on the other hand, gives me the kind of local results I get from Google.
Most local searches of that type tend to come from my phone, and this also helped me understand that better search on a phone matters WAY more than better search on desktop. The smaller screen and limited view means that it's significantly more important for the top result to be the one I want on my phone than it is on desktop. As such, Kagi is winning me over because it's made mobile searching frictionless -- something I couldn't say for DDG. That aspect alone is probably going to be what keeps me on the service. I'm planning on paying for at least another month, though after that I might go back to DDG for a month to see how I feel in comparison.
I mentioned earlier that I didn't think I could make it a default search on iOS. I mistakenly thought Apple had that locked down? Turns out it's actually possible through an app. Also, Kagi apparently has an entire browser for macOS/iOS. I tried it out and it works quite nicely, though AdGuard+Safari seemed to do a bit better ad-blocking than the stuff they'd built into Orion, so I've stayed on Safari.
There's actually a whole lot of cool looking power-user stuff on offer from Kagi (you can individually prioritize and de-prioritize specific domains across your searches, for example), but I'm not the kind of user that needs significant search depth, so I can't really speak to anything other than the standard search experience.
What I can say is that I've been very happy with that experience so far.
Also, it should hopefully go without saying, but this post isn't sponsored in any way nor was I requested to post it by Kagi. This is me choosing to give my own experiences with the service because I thought people here might be interested.
(Don't mind me shamelessly self-promoting.)
Whenever I see talk of search engines, one common thread seems to be that people want to use multiple search engines and don't want to settle for just one. Well, you can with DuckDuckGo's Bangs. You want to search IMDB?
!imdb everything everywhere. Wikipedia?!w. Google?!g. And the list goes on.And before you type that comment that says "but what about <website> that's not in the DDG Bangs list?" Fangs. Kagi as
!k? StartPage as!g? Do anything you want.Oh, and Fangs is also way faster than DuckDuckGo's Bangs since it doesn't have to go to a website to do its thing. It all happens locally in your browser. But that's just a free bonus.
At least on Firefox, you can also just create bookmarks that are keyworded and use
%sfor search parameters. For instance, with a bookmark likehttp://www.imdb.com/find?q=%s&s=allusing the keywordimdb, I can just typeimdb akira kurosawainto the address bar and search imdb directly.Well that would have been nice to know 7 months ago. :P
Chrome has a similar feature with tab-complete site searches. You can also add your own with the same syntax.
Interesting. I'm in the market for a new search engine because Google is so often choked with SEO junk. Sometimes DuckDuckGo is fine, but in my experience, when I'm searching for the answer to a specific programming bug, Google is much better than DDG. As a programmer, quality of search results matters a lot to me - the right link can save me hours of wasted effort. I'd be curious whether any programmers have tried using Kagi or Neeva.
I could just get into the habit of trying multiple search engines. You.com is a search engine that does that for you; I found their layout annoying last time I was there, but it looks like they've changed it since then, so maybe I should give it another try.
Yes, if you consider a physicist who does a fair amount of programming to be a programmer. I mostly use kagi: it has worked, in my view, at least as well as Google, and it has the additional advantage that you can deprioritize and block individual domains in the search engine itself (while client-side extensions exist to do this for Google, they simply can't work as well). Most programming SEO spam seems to be from a small number of domains, so this can be very helpful.
DDG, on the other hand, seemed horrible for programming spam, and was one of the reasons I gave up on using it. There are ridiculous, basic cases where, for example, simply searching for a Python standard library module will return spam above the official documentation.
DDG also now has fake github mirrors on page one results.... Why!?
Do you have technical details? Because to get local results comparable to Google, sounds like it is getting results from Google.
Edit: according to the homepage, it is getting results from other search engines. Now the question is, is this paid service better than Startpage or other search engine aggregators?
This is what their FAQ says:
The first paragraph kind of reads like word-salad to me, and it does sound like they're pulling from Google, which actually makes me like them a lot less? I was under the impression that they were mostly doing their own crawling and whatnot -- especially given their comments on costs:
That's rather surprising to me as well. One of the bigger draws of kagi for me was that it seems better than Google (and especially DDG) at not having search results that are overrun by SEO-spam sites. I had assumed that this was because they were making their own index.
It's not, I suppose, that surprising that they could do better here, even using the same indexes, through good processing and filtering. Many of the same spam sites show up again and again, often with the same domains: the paraphrase-real-sources sites like w3schools, the innumerable-bad-articles sites like geeksforgeeks or towardsdatascience that always seem a bit like they're computer-generated off a line or two of information fed to some algorithm, the scrape-and-republish sites trying to outrank Wikipedia and GitHub (eg, gitmemory, which was somehow successful in drowning out github in results), the obviously-computer-generated/assisted comparison and list sites that have no real original content. Filtering these out from results would not be that hard, and it is perhaps more surprising that Google and DDG fail so strikingly at doing so. Even where kagi fails by default, it allows you to deprioritize and block sites and types of sites, a feature that Google oddly lacks (and used to have, if I recall).
Kagi's argument, I think, is that this type of customization, and removal of low-quality views-and-advertising-supported sites, is fundamentally incompatible with the advertising-based business models of Google and DDG. And in a way, that they seem to be reasonably successful at being better while not using their own index perhaps gives evidence against a particular worry, which is that they're actually just doing better because those sorts of sites are targeting Google and Bing (and thus DDG), and if Kagi were to become large enough, they would fall victim to the same attacks.
That figure seems high.
I agree, but then again, I literally have no actual knowledge upon which to base that.
One thing I’ve tried to do over time is uncouple my pricing expectations from services that are subsidized by advertising. It’s what makes me okay with paying for Fastmail even though Gmail is free.
With something like that though, I can compare different providers to get a ballpark estimate of what’s reasonable. With Kagi, however, I have no point of reference whatsoever for what it costs to process searches because I don’t know of any other search engines that you have to pay for.
Does it actually cost them that much, or am I getting fleeced? I have no idea.
EDIT: Did some searching and it seems there is a competitor in the same space: Neeva. Their price is $5 per month.
I'm going to echo what I said in the other thread: I'm betting that amortising the fixed and mostly-fixed costs across a larger user base will bring down the unit cost a decent amount. The numbers in their blog post seemed more or less reasonable to me, but a lot of them won't grow too much even with 10x the usage.
On that basis, and because I'm often saying we need to learn to pay for services like this if we want to fix the incentives around them, I've put my money where my mouth is on this and I'll be interested to see how it plays out over the next month or so.
@kfwyre thanks for a great write up, as always! If you haven't done so already, I'd strongly recommend you send Kagi a the link to this thread - when I've worked with companies at a similar stage/scale to them, this kind of well thought out, constructive feedback has been like absolute gold dust.
The fact that the 50 search limit shaped your behaviour, for example, is the kind of thing that can kill signup conversion while being surprisingly hard for the business to spot. It may well turn out that they intended to be generous on time, but that actually a one week unlimited trial would work much better, for example.
I also haven't forgotten about the Google Photos alternative you mentioned in that previous thread! Still excited to try it, just need to find the time to set up the self-hosted version and start testing out the capabilities.
As an alternative perspective, I've been running a personal searx instance for a while, and find it pretty good. I do miss ddg instant answers and bangs, so still use ddg a bit
Pinging @pallas here, since I don't think topic-level pings work (though correct me if I'm wrong!). Feel free to mark this as off-topic.