"Blogspam" getting into Kagi results

ryanjamurphy

Steps to reproduce:
I'm occasionally getting blogspam in foremost positions in Kagi results.

E.g., a search for macos youtube app yielded a page from janetpanic.com in the third position, but after clicking through, it became clear that the site was an aggregator for all kinds of unrelated content in order to attract clicks.

Sample results for "macos youtube app"

The content on the offending page

Similarly, the query macos youtube playlist app returned heavyplate.weebly.com in a fairly high-ranking spot, but it was the same story:

Sample results for "macos youtube playlist app"

Expected behavior:

These kinds of predatory sites are hopefully something Kagi can get rid of over time—they're largely what has ruined Google search result quality, in my opinion.

Debug info:

Searching in Orion
"Canada" region setting
"US-EAST" Kagi server

Vlad

ryanjamurphy How do you propose we detect a result is blog spam?

gesumin

Maybe when clicking "Block" here kagi could ask you whether you want to report the site to the team, so that you guys could review it and possibly improve the results.

NunoCampos

Vlad

How about something like the non commercial results filter (nr of trackers etc) combined with volume of searches for the website name?

ie. if say CNN.com has lots of trackers but lots of people search specifically for “cnn” by name then you know it’s a legitimate website, therefore the nr of trackers can’t hurt its ranking in main results

however, if say, randomblogspam.com has lots of trackers AND the volume of searches for “randomblogspam” by name is very low** then it’s safe to consider it a low quality site that can demotes in results?

** (I think you’d need to normalise this “lowness” threshold by maybe the nr of different queries that pages of that website show up in)

ryanjamurphy

Obviously a tough question!

In each of these problematic results, it was clear that the page was a hodgepodge of random content. Each "block" of content had nothing to do with the other "blocks." Maybe there's a way of using the graph-theoretical nature of this to downrank potentially corrupt SEO efforts? Uprank results that have a higher concentration of links that are themselves more highly connected to each other, and downrank results whose connections are not connected to one another.

Does that make any sense?

Vlad

ryanjamurphy Any method used will have a counter method, Google has been trying to combat this for 20 years with billions of $ - without success

Also Kagi does not have an index of the entire web to analyze every page.

Whatever we use has to be simple and effective.

Vlad

NunoCampos The problem is that the site in question https://janet-panic.com does not have high number of ads/trackers. In fact it has low. It is just a bad site (aggregrator)

NunoCampos

Vlad ah well then sounds like nothing can be done on this one

enterprisey

If many users block a site, it should go in front of a human (multiple? volunteers?) for review, and they can choose to make it blocked for everyone by default. People can opt-out of getting those blocks, of course.

Vlad

enterprisey Sounds good in theory but creating that process (human reviewers?) is hard from organizational/expense standpoint. Also we are really trying to not be in the business of being 'arbiter of truth'.