Each week we find a new topic for our readers to learn about in our AI Education column.
A quick glance through your favorite article aggregation website or social media platform will instantly reveal that the internet looks quite a bit different now, in October 2025, than it did in 2015. Or 2020. Or even 2023.
And AI is the biggest reason why.
Welcome to AI Education, where this week we’re going to discuss—and attempt to fairly define—the term “AI slop.” If you look around, AI slop is everywhere. For some reason, it often involves the combination of something cute with something delicious, so, imagine a stylized image of a kitten eating a plate of powdered donuts, or an adorable litter of piglets chowing down on a pan of bacon. That’s kind of what AI slop looks like.
Put simply, AI slop is low-quality content produced (quickly) with generative artificial intelligence, however, it’s commonly used across the internet to describe any AI-generated content that a person has been involuntarily exposed to, no matter the quality.
In other words, if we were to go looking for discussions about AI slop, or initiate one ourselves, it wouldn’t take long for the conversation to be dominated by people complaining about any and every exposure to generative AI in their daily lives. Which, to us, means that there are people out there who are so anti-artificial intelligence that they’re never going to give the products of generative AI a fair shake. We’re not even going to bother to try to win those people over, so that’s not the kind of AI slop we’re going to talk about today—instead, we’ll talk about the huge amount of poor content being created with sophisticated generative AI that really deserves to be used better.
How We Ended Up with AI Slop
We arrived at this topic doing a little background for the AI & Finance column this week—we were originally going to write about open-weight AI, but that can wait for next week. AI slop is a neologism coined in 2022 in response to the launch of the first public-facing AI image generators, and rather than describing a reality, it was a prediction by technologists and technology enthusiasts that the advent of generative AI would lead to a glut of quickly generated, poorly edited filler content.
And so it did.
AI slop probably started in the social corners of the internet as a gag—as in, contests on social media and discussion communities to find out who can make generative AI spit out the most ridiculous picture. It may also have occurred by accident when people tried to “jailbreak” generative AI models—that is, create prompts that circumvent their controls and parameters to create provocative or problematic content.
Then the corner-cutters got ahold of generative AI—knowing that images were key to driving web traffic, people running scams or working for content farms on social media platforms and on the web used quickly generated to help get eyes on their material and pad their clicks. There was an upside—the same image generators allow smaller publishers without an art department of their own to create images to accompany their stories, leveling the playing field with entrenched publications commanding larger budgets. Then we all gained the ability to generate voice content, like podcasts, and later, video. Suddenly, AI slop was everywhere, and the perpetrators were both good guys and bad guys.
What Isn’t AI Slop—Spam AI
In our opinion, AI slop is not merely un-wanted AI. We don’t care for some applications of generative artificial intelligence that we’ve encountered. Sometimes it feels intrusive, particularly in a retail environment where we’re usually looking to make an easy purchase—that doesn’t mean that its product is slop.
On social media in particular, there is a vocal element that complains about clearly unrealistic, AI-generated images and videos merely because they don’t like AI. In many cases, these images and videos are in their own way striking, unique, provocative, emotionally stirring, though-provoking—you know, just like art. But something within the creative community rebels at the idea of AI-generated art. AI expressionist or surrealist paintings. AI pop bands. AI poetry. Not all of it bad, not all of it low-quality, but it all receives a visceral response from some online.
We’re not sure if they’re right or wrong, but we wish they would stop short of calling any and every product of generative AI slop, because a lot of what we see had to be created with intention. You don’t get to AI memes like “shrimp Jesus” or “Pope Francis in a puffer jacket” without carefully prompting an image generator, refining your prompt, and then taking some time to edit the image. That might not be art to all of us, but it’s not necessarily slop, either. It’s more like spam—some people don’t want it and think there’s too much of it, and they’re entitled to their opinion.
What Else Isn’t AI Slop—Misinformation
Misleading AI images are sometimes called AI slop. For example, during 2024’s Hurricane Helene an image of a cute but terrified little girl clutching an even-cuter puppy in a boat was widely circulated on social media—but that image was not from any area impacted by the storm, it was generated with artificial intelligence. Yet it was good enough to fool most people at first glance. That’s not slop.
It’s worse. It’s misinformation.
And that’s one of the dangers many technologists saw coming with AI, that it would become a tool of misinformation. One of the most popular AI-generated images to circulate on the internet was of Donald Trump’s arrest—an arrest that never happened. But the image was so realistic that it fooled some people. Was it slop? No, it was worse.
Not only were some poorly informed people taken in by such an image, but it was also circulated to fuel online hatred.
Wait, Don’t You Produce AI Slop?
Every week, we prompt an image generator—usually OpenAI’s DALL-E, to create art for AI Education and AI & Finance. Al & Ivy’s podcast is created using NotebookFM every week. So in some ways, we’re part of the problem, too, even though we try to carefully and intentionally write our prompts and at least provide a little bit of editing before publication. Our images aren’t produced at random, are appropriate to our pieces, and are very rarely re-used—most are unique to the story they accompany. Al & Ivy, week after week, are generated talking about a relevant topic using screened source material. They are not slop. More importantly, we don’t try to pass off AI as anything else, and we’re producing a newsletter and website about the technology itself, so it makes sense to employ it where possible.
Can we do better? Absolutely, and we will, and so will the AI tools that we use, and so will the AI tools available to us to help us screen out AI slop.
Speaking of the improvement of generative AI: The evolution of generative artificial intelligence tools moving forward might make it difficult to discern AI slop from everything else generated by AI—and that will open an entirely new can of worms. The time to get a grip on AI slop is now.
