1) What are the downsides to having your search history monetized? Is it the ads served to you across websites that use the same network?
2) Does personalized weighting lead to echo chambers like, say, Reddit?
I need to try this out but I'm curious about how they will defeat SEO. To me, that is what has made Google search results less helpful to me the past several years and would lead me to look elsewhere.
Beyond the obvious privacy concerns I'm not going to bother repeating, it really depends how it is monetised - is it only used to target ads on other products or does the search engine itself have ads?
The former option is potentially fine as long as you're happy with the privacy concerns. The latter option (which is what Google is) directly incentivises them to degrade search quality so that you spend more time on the search engine and potentially click an ad.
Furthermore, when it comes to a large company such as Google that also happens to control ads/analytics on many other sites, it incentivises them to preserve ad/analytics-infested sites in search results (since they make money and/or collect data off it) despite those being very good signals a site is spammy and should be downranked.
(for antitrust reasons they will never explicitly talk about prioritising sites with their own ads or explicitly degrading search experience - however the unspoken rule is that projects/ideas that might improve the experience and lead to less ad impressions overall might not succeed for various innocent-looking reasons, giving them a layer of plausible deniability)
A user-funded search engine doesn't have those misaligned incentives and can beat SEO by using common hallmarks of spammy content such as presence of ads, analytics, affiliate links and other marketing tools (used to measure the effectiveness of the spam) as negative ranking signals. This will be difficult to evade by the spam sites because the same techniques they'd need to evade search engine detection (such as self-hosting and proxying the ads) are common in ad-fraud and would be prohibited by their advertisers, thus cornering the spammers in a game where the only winning move is not to play.
I guess it depends on your threat model, but there are also privacy concerns with a search engine you have to make a paid account for.
For a regular user who signs into Google to use Gmail, they'll be searching while signed in anyway - no real loss of privacy, and arguably much gained by not broadcasting anything into the adtech data pools.
For a very privacy conscious user who wants to be anonymous and is already taking steps like using VPNs, there is a loss in privacy due to payment info being attached. (You could argue Google accounts could be pseudo anonymous, but for most users they're anything but - location data, history, emails and search queries make them totally identified).
There are technical solutions to avoid the payment linking issue (cash in envelope, selling gift cards via retail, etc), or through using blind signature schemes to separate payment from usage. The user experience of these options is likely to be far more complex, with people losing access and customer service fiascos with people being unable to prove they signed up before etc.
Which is an option with Brave and DDG, but obviously not Kagi. Yes, obviously, this is only going to be a relatively fringe use-case that depends on the user having generally decent opsec.
No need to argue the abstract; all that matters is the specific. If Kagi specifically is abusing trust, I’d love to hear it.
If Kagi is being honest, typical usage of Kagi is probably more private than typical usage of advertising-supported search engines. But I can more easily verify that using e.g. Brave Search via Tor is private.
True. But even the payment thing could be solved although most users won't care. Mullvad comes to mind: you can send them cash via mail and the payment is associated to a serial number that is simultaneously your login and password.
That's not really the point. It is up to Kagi to prove to their customers that they are trustworthy, not up to the customers to find abuses of trust.
If I -- for example -- use Tor and a private browsing window to access Google search, I have a pretty good expectation of privacy. I don't have to trust Google here; they simply cannot correlate my search with my Google account in this case, at least not 100% reliably (browser fingerprinting, etc.).
But with Kagi, they know that every search I make, regardless of how I make it, is from me (or, at least, from someone using my account), and that account is tied to my real-world identity. Sure, they could implement a more complicated payment method, where I pay a third party, and that third party blind-signs some credentials that Kagi trusts. But they haven't done that, at least as an option, and that bothers me.
When it comes to third parties, the best kind of trust is the trust that you don't have to have. I have to trust that Kagi isn't keeping a record of my search results. I would rather that I just don't have to trust anything, because there's nothing nefarious they can do.
Fair enough. Since that is a provably impossible bar, you’re basically saying nobody should do business with anyone. Which I admire from an ideological purity perspective, but don’t really have time for at this age.
If I'm doing something in private browsing mode, I fall back to DDG. Not only do I feel weird about being signed into something when I'm supposedly doing something "private", either I have to manually sign into Kagi each time I open a private browsing window, or install their extension, neither of which I really want to do.
So I'm torn; I'm not sure if I'll keep paying for the account, or cancel it and go back to DDG.
Then again, I also pay Fastmail for email, which gives them access to all my email. Arguably that's just as bad (if not worse) for my privacy than giving someone my search history. It's not clear to me why I'm ok with that, but uncomfortable when it comes to Kagi. I suppose email is inherently tied to me, whereas search results need not be? I dunno.
Regardless, this is the big issue with paying for something on the internet that you can also get for free. In some ways my search privacy can be "trustably" better with Google (via private browsing + Tor or at least a VPN) than it can ever be with Kagi, which is weird.
2. Personalized weighting just means I can downweight websites I know to be not what I'm looking for, like Pinterest, and I can upweight websites that probably are more like what I want -- stackoverflow, wikipedia, etc.
This is the best, and frankly the only good argument I know. The privacy concerns are IMO overblown, but the skewed incentives are the main reason why we ended up with internet that is toxic and miserable.
One of the main reasons I use Kagi is similar to why I use Firefox & Brave. I want to encourage competitors so that innovation continues to happen.
But a search business where you pay for search is more likely to be privacy-respecting than a search business where a 3rd party (advertiser) is paying for your searches, just because the business model aligns incentives.
2) or HN? :-)
I also just don’t want to see tons of advertisements pretending to be results and figuring out which is which. Google also seems to be particularly susceptible to SEO garbage. Between their own ads pretending to be relevant and SEO garbage pretending to be relevant, I find myself scrolling through pages upon pages of advertisements hoping to find some signal. With Kagi I spend much less time discerning between garbage trying to capture my attention and something about what I’m looking for.
(I’d note here, as a bit of a non sequitur, that if/when something like chatgpt becomes more accurate and can provide references and do information retrieval better, I’ll be happier to use that at an even higher price point than Kagi. I find the ability to ask a question and get an answer even better than searching for documents that contain symbols in my term and personally assessing the contents for clues about an answer to my actual question. ChatGPT reminds me of the moment I clicked “I’m feeling lucky” on some webpage someone sent me in 1998 - the tedium of answering questions had reduced dramatically.
That said, I think the semantic understanding required to answer questions in the ChatGPT way will allow for even more nuanced personalization, and I’m afraid even more insidious (1)’s.)
Lots of people value privacy in and of itself. If you don’t care about Google recording and mining your search history, then a core value proposition of Kagi doesn’t apply to you. Kagi is targeting the subset of the market that cares enough about privacy to pay for it.