Add ability to report sources as slop in Assistant

tapeless839

When using Assistant with search enabled the model will be fed a source that is obvious AI slop, but there is no way to report this source through Assistant UI. I'd like to see a report option next to each entry in the References section (and/or the sources in the thinking output).

Even better would be integrating the entire domainInfoModal (shield icon) component from Search. The user would be able to see the same information about a site before choosing to navigate to it. This would also allow up/downranking, making it clear that these settings affect both direct search results and Assistant's sources (per #9229).

MustafaD

Same should apply to videos which appear in Web Search, as reported at #10302.

Pleonasm

I agree that adding the ability to report AI slop within Kagi Assistant would be valuable.

Yet, from a larger perspective, I wonder: what is the goal? Is it reasonable to believe that Kagi, with its limited number of users and searches per day, can realistically create (and maintain) an AI slop curiated blacklist that meaningfully improves the quality of Kagi Search results?

If the goal is not to create a blacklist, but instead to create a predictive model that calculates the probability that a webpage is AI slop, then reporting AI slop by Kagi users helps to generate a model development training dataset. However, Kagi has yet to disclose the extent to which this predictive approach is demonstrating early indications of success, and how it is (hopefully) overcoming significant challenges.

MustafaD

Pleonasm I think it is based on the principle of trust.

That is, a malicious user can report valuable and human-generated sources, negatively affecting service for everyone.

Then, a large user base increases the appearance (not probability) of malicious users.

But because Kagi is a paid service, with a relatively small user base, it has the advantage of such a simple feature being useful to everyone.