SlopStop Report Feedback

fanged_bagful

I reported the following article https://oreateai.com/blog/choosing-between-14-and-16-gauge-extension-cords-what-you-need-to-know/a7b2d2e7798e848406b5b8bba0aa35c5 as ai slop.

I was looking for an extension cord to use for my space heater and was looking for information to make a safe choice. Once i realized that that the article was ai-generated, i decided none of the information could be relied on to make such a safety decision. So i reported it as AI Slop. I just checked at the report from 2026-02-25 was marked as status "Not Slop".

I expected it to be marked as slop, because the article link clearly reads as ai-generated with the way it breaks down the information and summarizes it. Further, the blog for that domain is filled with a huge amount of blog posts of very varied topics that all look ai-generated. Also, the domain itself advertises an AI service that i assume is reposting articles generated by its users into the blog of the domain. I found this article in a Kagi Search, so this domain is being succesful in SEO as a way to advertise their service. By including all these random articles in their blog, they are able to surface on a large amount of queries.

Further, I would like to know what the background processes were that led to this specific slopstop report to be marked as "Not Slop". Because it seems like it has a long way for the program to mature, in order for it to be properly marking ai-generated content if it led to this pretty egregious false negative.

fanged_bagful

The following are additional links that are sub-listed under the orginal search result that is reported. All of them are appear to be ai-generated content using similar prompts.
The first three articles,if the blogs timestamp is accurate, were posted on the same day Jan 27 within 2 hours of each other. What signals is Slop-stop relying on that it missed this? Also, i found the shield button only lets a user report the first article listed in the results and not the related articles that are sub-listed.

https://www.oreateai.com/blog/14-gauge-vs-16-gauge-extension-cords-whats-the-real-difference-for-your-power-needs/6cf89ee96991360a956c02de5ed4b06c

https://www.oreateai.com/blog/16-gauge-vs-14-gauge-extension-cords-whats-the-real-difference-for-your-power-needs/4da98fc7e5fe00df8c93c2101df46ae0

https://www.oreateai.com/blog/beyond-the-numbers-why-16-vs-14-gauge-matters-for-your-extension-cords/d4570e19feefb3e93b4a0da5973e5e9a

https://www.oreateai.com/blog/beyond-the-gauge-understanding-the-real-difference-between-143-and-163-extension-cords/ce2d47f3cfac14894fd23defd0704faf

https://www.oreateai.com/blog/powering-your-outdoor-projects-a-guide-to-choosing-the-right-extension-cord/b45374abd255cb66c3493c004da65fef

k7absp2h

It's 100% slop

RonanCJ

There's some ongoing discussion and a dev comment in this thread I started. You may want to post your links there as well @fanged_bagful

k7absp2h

RonanCJ Thanks, but until a public solution is released, I don't want to make any contributions.

fanged_bagful

RonanCJ Thanks for linking to that thread. I thought this StopSlop program would maybe have trouble with more disguised AI-generated sites, but to miss on these pretty obvious sites shows that something in their process is really deficient.

They had us submitting reports for several months before going live with processing all the reviews, i thought they would have worked out the kinks and any edge cases by now.

1player

Same issue here. I checked today, and two sites got marked as non-slop:

https://microwaveshub.com/ which contains 3,121 pages of posts about microwaves
https://cyberpost.co/ which contains 6,185 pages of posts about videogames. It is only a few hours since midnight March 1st, 2026, and there are already 30 pages of posts with this date.

If this isn't slop, I don't know what is.

RonanCJ

Two obvious AI generated websites were reported by me, but when reviewed by Kagi, they were marked as "not slop".

I expected that, as a paid search engine with the stated mission of "humanizing the web", Kagi's Stop Slop service would identify and downrank slop that is obvious to a human. The first website has several of the signals previously mentioned by Kagi: over ten thousand articles (far beyond what a human could write) with no images, a basic website theme, no sorting/"classification" in the URL, and long repeating prose that tries to cover all possible search queries. The second website is somewhat more elaborate, but still has some obvious signals: an impossible update frequency, long repeating prose for SEO, a relative lack of pictures, and a table of contents template identified by another Kagi user. (Hat tip @fxgn )

Whatever system you're using to automatically classify slop will not work at scale if it lets through low quality pages like this.

E: This thread can probably serve as one common place for other customers to list websites they reported as slop but got classified as "not slop".

fxgn

https://thisvsthat.io/spectrometry-vs-spectroscopy

This is very obviously AI generated SEO slop. If you check the homepage of a website, it lets you type any two things to compare them. It is obviously not human written and I have no idea how Kagi moderation marked this as not slop

mxm

So far, every obviously AI generated page I've submitted has been marked "not slop" as well. I hope this gets fixed

k7absp2h

Kagi SlopStop is a marketing lie; they don't do anything about slop. Even open-source community slop filter lists are so much better (e.g., Super-SEO-Spam-Suppressor or BadWebsiteBlocklist).

My 3-month-old submissions, e.g linuxvox.com, is still there (btw, open-source filter lists filter that).

I can't understand. You have the best search-aware user base, but you still suck. It's unbelievable.

Just make user submissions public and have an active developer on that.

fxgn

k7absp2h My 3-month-old submissions, e.g linuxvox.com, is still there (btw, open-source filter lists filter that).

They only started reviewing submissions this week, so that's understandable. The issue is with the sites they do review and explicitly mark as not slop when it clearly is

k7absp2h

fxgn But still, why can't we see the process?

nobodywasishere

I apologize for the delay in evaluating your reports. While not perfect, for those we've evaluated so far, >98% of people's reports were slop, so these cases are important for us to handle to make sure we improve our systems for catching slop. If y'all post those examples here or re-report, we can take another look at those.

k7absp2h We have been planning on doing a more thorough deep dive into how SlopStop works in a blog post. We can take a look at the open source lists you posted to see about integrating those as well (we already integrate some others too).

jagotu

This was also maked as "not slop": https://www.codestudy.net/blog/nginx-as-cache-proxy-not-caching-anything/

Section 3.3 contains an obvious hallucination, "proxy_cache_log" directive does not exist and has 0 search results

k7absp2h

nobodywasishere Thanks, but you don't have to reinvent the wheel from the beginning; a simple repo could be good.

And I assume all old requests are evaluated, right?

If so, please clarify why linuxvox.com is not slop?

kRush-

My report simply vanished after ~3 months. The domain still turns up in search - no downranking or indication on the shield:
https://www.isteshaltbar.de/frage-und-antwort/?page=40747
If content-farm crap like this flies, what's the entire point?

nobodywasishere

kRush- Thank you for the link - I can take a look at it.

fanged_bagful

nobodywasishere

Hi, Could you please take a look into this report?

If more than 98% of all reports submitted were marked as slop, and thus only 2% were marked as "not slop". I wonder what poportion of that 2% were false negatives like the one i posted and others in this thread have mentioned.