I saw a thread on an SEO subreddit this week, and it’s a question I’ve been seeing more and more. Someone asked: does AI slop work?
The simple answer is yes. It works.
It’s just more complicated than that.
What Slop Actually Is
First, a definition, because this note is easy to misread as “AI content is bad,” and that’s not what I’m saying.
AI slop is AI content generated at scale, or with almost no effort. Very little human input, no editing, no review. Generate it, post it, generate the next thousand.
The tool isn’t the problem. Google’s spam policy on scaled content abuse is explicit about that. It describes the problem as many pages generated “for the primary purpose of manipulating search rankings and not helping users,” and says it applies “no matter how it’s created.” AI is named as one way to do it, alongside scraping and stitching other people’s content together. The pattern is what gets targeted: lots of pages, little value, nobody home.
AI-assisted content that a person actually reviews, edits, and stands behind is a different thing. Last week’s note on fact-checking AI output was about exactly that kind of review. Slop is what you get when that step doesn’t happen at all.
Yes, It Ranks
Here’s the honest part. Slop can rank, and sometimes it ranks very well.
But what we’re seeing is that it doesn’t last.
Watch any SEO forum right after a Google spam update or core update rolls out. There’s a flood of posts from people saying their site got hit. And the most common thread running through them is AI content generated at scale. Thousands of pages. A fair amount of clicks, tons of impressions, and then, all of a sudden, zero or near zero.
Google doesn’t publish exactly what each spam update targeted, so you won’t find an official line saying “this one was for AI slop.” But you don’t need one. The people getting hit are telling you.
Google Is Telling You, Too
You also don’t have to guess where this is going, because Google keeps saying it.
Using generative AI to produce many pages without adding value is listed by name as scaled content abuse in Google’s spam policies. Google’s guidance on AI-generated content repeats the warning, and that page was updated just this month.
And look at the pace. Google’s Search Status Dashboard shows four spam updates so far in 2026: March, June, August, and September. In all of 2025, there was one.
That’s not a company that’s undecided about the problem.
We’ve Seen This Before
This reminds me a lot of the stretch just before Penguin launched in 2012.
Back then, I used to get into arguments on forums with people using really low-quality links. And they were getting great results. Profile links, bookmarking links, blog networks, all of it was working. Telling them it wouldn’t last sounded like jealousy.
But if you were paying attention to Google, the signs were obvious.
Google sent more than 700,000 manual action messages to site owners in the first two months of 2012. Most of those weren’t about links, but the link warnings were in there, and Google was clearly building toward something.
Then it started going after the link networks themselves, one at a time. On March 19, 2012, Google deindexed the overwhelming majority of BuildMyRank, one of the most popular networks at the time. BuildMyRank shut down the service and offered refunds.
It didn’t stop there. I watched it happen to Authority Link Network. Same thing. I watched it happen to SEO Link Monster, which had only just launched. Google went after them one by one.
Then, on April 24, 2012, about five weeks after BuildMyRank went down, Penguin hit. Google said it affected about 3.1% of English queries, and a huge share of the sites that had been winning with low-quality links got leveled.
People had their businesses devastated. Not because nobody warned them. Because the tactic was working right up until the day it wasn’t, and the warning signs were easy to wave off while the traffic was still coming in.
Getting Back Is the Expensive Part
The drop is the part everybody talks about. The part nobody prices in is getting back.
With Penguin, many hit sites couldn’t recover until Google ran the next Penguin refresh, and you had no say in when that would be. Penguin 3.0 ran in October 2014. The next one, Penguin 4.0, didn’t arrive until September 2016, when Penguin finally became part of the real-time algorithm. Sites hit in 2014 waited nearly two years for a chance to recover, even if they cleaned everything up the week after.
Getting hit is fast. Getting back is slow, and sometimes it doesn’t happen at all.
That’s the math the churn-and-burn crowd doesn’t do. A few months of gains against a hit that can take years to recover from, if it recovers.
Slop Doesn’t Stay in Its Corner
There’s one more cost, and it’s the one that should worry people with real businesses the most.
A few weeks ago, in the note on deleting old content, I quoted Google’s 2022 guidance: content on sites with “relatively high amounts of unhelpful content overall is less likely to perform well in Search,” and that includes all the content on the site, not just the unhelpful pages. The standalone system that guidance described has since been folded into Google’s core ranking systems, but the idea that a site’s overall quality affects every page on it survived.
So if you publish a thousand pages of slop on the same domain as the pages that actually make you money, you’re not just risking the slop. You’re risking the good pages too.
And a subfolder isn’t a firewall. Putting the slop in /blog/ or /resources/ doesn’t wall it off from the rest of the site. It’s still your site.
What I’d Recommend
If you’re running a churn-and-burn project, a site you don’t care about, one you’re fine watching get hit and never recover, and you’re just trying to rank something and make a few bucks quickly: yeah, AI slop can work. It might work for a few weeks. It might work for a few months. It might work for close to a year. But at some point it’s going to get hit, and you should plan around that day, not hope it skips you.
If your website is a significant portion of your income, or your only source of income, stay away from it. Don’t start generating AI content at scale.
And whichever one you are, don’t mix them. Never put slop on the domain that pays your bills, not even a little, not even in a subfolder, not even as a test.
AI slop is easy, and that’s what makes it attractive.
But it’s going to get hit at some point. It’s just a matter of when.
