In the recent days I’ve been stumbling upon weird, new so-called “AI” Mathy-math-slop sites, like linuxv*x.com[1]. Some other was called something like “tutorialsipedia”, or whatever.
Have you noticed these? Is that some weird new Startup that wants to leverage CEO and “AI”? I’d use them, but my eyes glaze off the page. It’s like a drop on a Lotus leaf and I can’t really read that garbage. What’s up with those?
Don’t want to give them the traffic. ↩︎
That’s just how search engines go now.
Lately I’ve been seeing websites that steal content from Stack Overflow verbatim rank higher than the same SO page itself.EU needs laws that AI-generated content must be marked as such. The US needs them too, but the fraudsters in chief there want you to be misinformed.
There’s AI slop websites for everything now. Flood the zone with enough garbage that traditional search engines can’t find anything, force people to use AI where the narrative is controlled.
And Ai tools are scraping the web… good lord.
I noticed this even before the sloppening I think it’s good old fashioned SEO
SEO-based business models used blogspam before. It’s the same SEO garbage that gets it into search results, but the content is now AI slop instead of contracted labour at pennies/word.
And search is garbage, now, because of enshittification; Google gets more money when you give up and couch the sponsored links, and re-query or load more pages of results to load more ads. So there’s no incentive for them to filter the spam.
Yeah, I think that’s a sad part of AI, is that even when you separate all the specific problems it has (energy use, CSAM, fake news) it also revivifies a bunch of old scams, SEO, phishing, etc
A super intelligent AI is also going to be better at scamming people than any human
The Sloppening. Hehehe.
The text is definetly slop, though.
It’s happening to all difficult problems. I’ve been searching for help with car wiring diagram or trouble codes and getting endless copies of slop scraped websites from DDG.
They often appear to be generated on the fly and rarely have any real information in them past the relevant search term. No real info. Super fucking frustrating.
TIL about footnotes in markdown. Neat!
Not working on voyager. ;_;
Wait how the fu…
Looks like markdown is converted to html syntax. In-text:
<sup class="footnote-ref"><a href="#fn1" id="fnref1">[1]</a></sup>And footer section:
<hr class="footnotes-sep"> <section class="footnotes"> <ol class="footnotes-list"> <li id="fn1" class="footnote-item"><p dir="auto">Don’t want to give them the traffic. <a href="#fnref1" class="footnote-backref">↩︎</a></p> </li> </ol> </section>From there they can be stylized. Pretty neat. More info
It’s kind-of funny. Nowadays, I find the AI search assistants (I used the one with Kagi) work better than search results with all of these shitty AI sites.
We’re back to the age of pre-StackOverflow, when Expert Sex Change was always plaguing my search results with fucking pay-to-view bullshit. Except it’s free-but-useless websites now.
This will change your life: https://ublacklist.github.io/docs
I use this userscript (works on Ddg, Bing, … too) since i can’t have a addon for every little thing. It can export/import the lists too.
was about to add this. kinda sad that I had to give up my searxng instance because I can’t get ublacklist to work on it (works on some public instances)
Neat… Is there an alternative for people who don’t use google? 😅
It works on each of these:
https://github.com/iorate/ublacklist#supported-search-engines
Searxng and ddg? Nice! 🤩🤩
Thanks a bunch! ❤️
It blocks on all the major engines. DDG included.
Awesome. How have I never used this before
Nature is healing.
How did I not know about this? Thanks!
OH man, thank you for that!
Wait, doesn’t this just duplicate uBlock functionality? You can just add the lists directly to uBlock, or am I missing something?
I have heard before that you can just add it to uBlock Origin, yeah.
Hey so just because I went on a spree, here’s a bunch of lists you can choose from to import into uBlock since that’s super simple already, and this other addon requires you to add them anyway so:
Thank you very much, do you have recommendations on lists to subscribe to?
Nothing in particular. They all seem about the same to me. Check top rankings on GitHub perhaps.
Lookup “splog”.
Welcome to the year of the linux desktop. Now solving linux problems is big business!
What you’re saying about drops on a lotus leaf hits though. There’s something weird about the prose on those sites that’s significantly different than even ai text I’ve made at home on my own hardware.
Sometimes it feels like the opposite of meditation where I can feel something tugging “up” in the top center of my skull when “reading” one of those pages but don’t remember what the page was about.
drops on a lotus leaf
Here’s a strategy for scoring your own search results.
“Keywords” are the seven words most commonly occurring on the page. If these seven words are seen to be repeated on the page to an unusual degree, then it is a good assumption that the page was designed by the author to appear high on search results.
Keyword density is a measure of “gloss.” Most people will read pages with high keyword density as unusually glossy. Keyword density is not necessarily related to how genuine the page content appears to be otherwise, but most people will look askance at a page that is too glossy.
It should come as no big surprise that the pages that appear high on search results have been designed that way. They are deliberately glossy with high keyword density. You may consider whether to skip reading them or even loading them in your browser. Chances are good that the glossy pages are mostly advertising.
Generally you will find interspersed in your results a handful of sites with low keyword density. These are likely from universities, government sites, and research institutions that have sources of revenue beyond advertising. You may consider whether to load these up and skim through them. Probably they will show a publication date, author, and list of references, which will move your research forward.
It can be noted that AI-generated sites often exhibit high keyword density. This is probably deliberate so that they garner advertising revenue. However, it may also be due to “bot 'splaining,” which is polly-paraphrasing a series of several (perhaps contradictory) articles.
Keyword density is not the only measure of gloss. There are others that have been developed to measure ratios between parts of speech. Unfortunately NONE OF THESE — including keyword density — distinguish sharply between pages that naturally convey genuine information and pages that have been designed to convey fluff for ulterior purposes. It is unlikely that combining measures of gloss will result in a tool that discriminates much better than keyword density by itself.
-
Piskorski, Jakub, Marcin Sydow, and Weiss Weiss. “Exploring Linguistic Features for Web Spam Detection: A Preliminary Study.” Airweb '08: Proceedings of the 4th International Workshop on Adversarial Information Retrieval on the Web. Ed. Carlos Castillo, Kumar Chellapilla, and Dennis Fetterly. New York: ACM, Apr. 2008. 25-28. ISBN:9781605581590. DOI:10.1145/1451983. 09 Nov. 2025 https://users.pja.edu.pl/~msyd/lingFeat08draft.pdf.
Nevertheless, you may wish to explore keyword density as a means to rank search results.
When I try to include a direct link to my python scripts, which do that, my responses and in fact the whole posted discussion are taken down. … something to do with self promotion of untested software I suppose. But you can find them in the Cheese Shop (See Wikipedia “Python Package Index.”) under clanker_score.
We don’t want to make this too easy for just anyone to censor all his search results. Rather, these scrips are meant as a learning tool. They demonstrate generally how rotten search results can be on one particular and not very compelling dimension. It should not be necessary to download and scan each and every page. You should be able to train yourself to ignore a priori results that include handfuls of pages from unauthoritative sites.
This is assault
-
“reading” one of those pages but don’t remember what the page was about.
That’s one of the biggest tells of AI-written text. It uses a lot of words to say very little, but does so in a very authoritative-sounding (or needlessly flowery) way.
Sometimes it feels like the opposite of meditation where I can feel something tugging “up” in the top center of my skull when “reading” one of those pages but don’t remember what the page was about.
This is your brain on slop.
You know, I was like, wtf is slop-y. If they mean it’s slop, it should be sloppy. But then I figured, sloppy is ambiguous, ok, so what about slopp-y, for clarity? But that makes no sense because, well, it don’t, so, all right, maybe slop-y, that would wo- ooooooooooooh…
I vote for “slopesque”, even if it has more letters. It doesn’t hurt that the most common English word that uses the -esque ending is “grotesque”, which this whole phenomenon is.
deleted by creator
