A page for a list of sites that are blocked/preferred by lots of users

RGBCube

Basically the title. There should be 2 pages, one "promoted", and one "blocked" those would have a list of sites that lots of people have blocked/promoted (sorted by the amount of people that have b/p ed that site). There should also be a button to apply the setting to your own account, if you are logged in. Filtering based on region, regex, etc would also be quite useful

Vlad

RGBCube We should have this. Where you imagine it would be found? Any idea on the interface?

NoGoogle

RGBCube How similar is the implementation of this idea to something like this? https://kagifeedback.org/d/802-lets-see-a-dedicated-page-to-what-fellow-kagi-users-have-boostedpined-so-that-we-can-surf-the-webs-greatest-sites

Recast

This is an interesting idea. I only block sites that are those AI-generated Q/A sites that have been springing up recently, since I don't trust them to provide good information. Obviously I can't block a good proportion of sites alone - but crowdsourcing such a thing via a shared block-list may be viable. c.f. sponsorblock's success.

Recast

As a user, I would expect to see this on the "personalised settings" page https://kagi.com/settings?p=user_ranked&k=1 with a button next to "add domains" like "popular domain rankings" to get into the interface. I would expect the interface to show me a table of domains with the number of votes and a quick action to copy them into personal settings.

This placement would be fairly hidden - maybe you'd want to hint at this being a feature under the crystal ball - but I suspect many of your users are proficient with technology and have interest in browsing the settings pages, so might not be an issue.

Vlad

Recast That sounds like a plan

NoGoogle

Kagi's boost feature is unintentionally the greatest web surfing tool that I know of. Kagi should have a special page dedicated to showing only the collective domains that Kagi users have boosted/pinned with a number count next to each domain showing how many Kagi users have boosted that same site ordered from greatest to least. As Kagi user's, we are boosting sites we love so much that we are willing to see them in our search results over and over again. That level of love for a website is one of the most organic ways of showing which websites are genuinely the most valuable/popular to users on the internet. It would be the quickest and greatest ways of finding the best content the web has to offer all in one central place.

Doing a Kagi search for "best websites the internet has to offer" is not only subjective in the sense that it's coming from one person's blogging opinion but also highly manipulable by having payed sponsored sites as those being listed. Kagi unintentionally created an organic gold mine of genuinely liked websites updated in real time. Anyone who enjoys surfing the web for great sites knows it can be time consuming to find lots of great ones so Kagi having the ability to list them all on one easy to find up to date location is just extraordinary. To make it more inciting, browsing this these domains could be a premium feature only viewable to paying Kagi users. Since there isn't really an expense to maintaining this list, it would be a non resource taxing way of attracting more premium subscribers. Not necessary, but you could also exclude domains that have been boosted/pinned by only just 1 Kagi user by limiting the list to at least 2 boosts/pins or more that are allowed to show up in the event some site some Kagi user boosted was personal. The location of this page could be linked at the bottom bar of Kagi alongside the other linked URLs.

Vlad

NoGoogle If we did this in an automated way, it could potentially have privacy issues. So we would need to make it an opt-in mechanism, which would drastically reduce its usefulness. Thoughts?

seventhwave

Hello Vlad ,
I think that the "immediate monetization" for Kagi could be a double-edged sword.
In an ideal world, I'd like to see happen the NoGoogle proposal, but my biggest fear is that site could become to pay to be listed into this set.
This could lead to a perception of lower quality: I think that distrust in the community would mine the overall Kagi perception by the standard user.

That said, if you're confident about this idea, a possible implementation could be:

the first time that a user boosts/pin a website, the user could be asked to share this boosts in an anonymized way or not (this setting must remain available in global settings)
in settings we could add a switch about "highlight boosted results" (with an icon or a border)
(and yes, only to Kagi subscriber)

About the "opt-in mechanism, which would drastically reduce its usefulness": I think that, once activated this setting and seen the positive net result on their searches, lots of users will enable the boost sharing .

What do you think?

NoGoogle

Vlad I think the concept of a lower limit is the privacy preservation in itself. Up the limit to increase certainty. If say 50+ Kagi users have boosted the same site, it can be reasonably safe to to say that the website in question is publicly popular enough to not be something that need's to be kept secret especially considering that websites themselves are publicly accessible. So no to the opt-in mechanism as long as a good minimum to be shown publicly exists.

To rephrase, this suggestion is just a dedicated page found somewhere like Kagi.com/TopRanks that is bare bones simple and extremely useful. It would look very similar to this.
Kagi.com/TopRanks webpage is separated into two sides. On the left side it shows exactly like this:

Kagi's Favorite Boosted Sites
News.ycombinator.com (1043 Boosts)
MakeUseOf.com (543 Boosts)

On the right side of the page it shows Blocked domains:

Kagi's Most Hated Sites
CNN.com (1051 Blocks)
FoxNews.com (953 Blocks)

As you can see no info from who those 1051 users who blocked a specific site are.

RGBCube

Vlad It could be only shown if a lot of people have done the same thing. So it wouldnt be a concern (as @NoGoogle has said in the previous replies)

Vlad

seventhwave So the user is asked if they want to participate in global personalization scheme meaning

their own domain personalization would contribute towards towards global personalization
they would get personalization based on global database

So how does this further work?

Is one count enough? How many is?
What if a domain has opposite signals?
Are global preferences visible somewhere?

NoGoogle

Vlad I think I see two different ideas here. Mine being just a nice way to see what others are boosting and blocking and another interpretation of it to use the websites Kagi users are blocking and boosting to further improve Kagi's own default ranking algorithm. Two ideas that I'd LOVE see become reality.

I wish I had a link but I was reading from somewhere where Google once did an internal test by letting some users decide the PageRank algorithm for themselves. Websites that users loved a lot were upvoted and websites that weren't were down-voted and it organically led to reduced SEO spam better than what Google employee's internally could keep up with combating and had superior results to their default ranking algorithm (if true then a good reason why they stopped it). It's a powerful concept, to let those using the search engine direct it's improvement instead of just a smaller group of workers at Google working 24/7 struggling to keep up with the forever constant influx of new SEO spam off from the search results.

As for further details on implementation, the more users that raise a website, the greater weight that website should have in Kagi's default ranking algorithm. Same for lowering and blocking but with sites blocked having 2x or more the weight as lowering a website. You're pooling manpower together to further refine the web. Right now we are all in our own little bubbles blocking the same sites, boosting the same cool sites but with no way to pool our efforts together to further refine and build upon the efforts of others. Doing so would be like having AI robot's that once one makes a mistake, all others are aware of it and know not to repeat it, where one learns a skill all others have it as well but with this ability being applied and used in refining search results.

greenbluepig

Vlad It should be a mode. Like single vs multiplayer. That's straight forward and the privacy implications are self explanatory with that dichotomy. If you're in single player mode your block list doesn't get shared. If in multiplayer, you are in a shared list. Kagi Custom vs Kagi Team, etc.

seventhwave

(I'm continuing to brainstorming hoping to help us)

opposite signals: do we need to consider all choices block/lower/higher/pin? My idea is that: with this scheme we want to convoy to the users that "there are some choices that lots of users already like".
I think that the extremes (block/pin) could be used as a strong personal preference, modeled around the individual story (are you a british metereologist? maybe you want https://www.metoffice.gov.uk/ always on top; do you scream when you look at rats? you'll block http://www.ratbehavior.org/, you don't want to see this site anymore)
About lower or higher? I think that lower could be like "for this search, this website is less interesting than Kagi told me"...and I think that later, the users that will see the "lowering", will not feel a sense of community or connection, but only something like "this is a bad result from Kagi".
So, I'll select only higher. You'll can further think about using the "lower" votes for analyzing your result quality (but personally I think it will be a dangerous road)
how many counts?
I think that the only correct solution is that has to be proportional to the total "higher" number in the user panel that partecipate in this scheme. If you've 100 users partecipating and 350 "higher", maybe 2 or 3 "higher" are right. But what about having 5000 users with 12000 "higher"? if half of the results are marked as "liked", this isn't useful anymore, and quickly become noise.
global preferences
yes I'm thinking about having this preference on global settings: you must be able to change your idea later

sounds useful? do you like to elaborate further?

rrwoods

Here at the start of 2025 wishing for something like this. I would like it if known AI slop domains got downranked.

As a shining (but certainly not the only) example, goes into how several search engines currently give essentially an entire page of AI slop/marketing for a simple search of "glb file format". Kagi unfortunately returns many of the same sites presented in this video. I would love it if Kagi could dedicate some resources to finding and auto-blocking (or severely downranking) these sites.