If you’ve spent any time on Facebook recently you’ve seen the river of AI slop. There’s the homeless vet with the badly worded sign, cops carrying massive bibles, and—of course—Shrimp Jesus. Why is Facebook full of this stuff? Because there’s money in it, of course.

A new investigation from 404 Media dug into the origins of Facebook’s AI slop era and uncovered an ecosystem of YouTube videos, Telegram channels, and reel-filled Google drives. It exists to automate the creation and distribution of AI-generated images on Facebook.

People in places like Pakistan, India, Vietnam, Thailand, and Indonesia use tools like Microsoft’s AI Image Creator to churn out content. They post dozens of images every day, sometimes every hour, and make a living from the engagement.

The goal is to get up as many posts in a day across as many accounts as possible. An account that’s consistently posting can get invited into Facebook’s Creator Bonus Program and start making money from viral content. Some posts will only generate a few cents while others may pull in hundreds of dollars. The more posts, the greater the returns. The goal is to use automation to post as much as possible, sometimes hundreds of times a day.

Like all spam circles, there’s a whole other economy filled with folks who will sell you the tools and knowledge you need to get started. There are YouTube channels dedicated to getting people’s AI slop machine up and running, dozens of guides on Fiverr and Gumroad, and Telegram channels full of simple English prompts that will help the AI image generator create a viral picture.

Some creators 404 Media uncovered were running multiple Facebook accounts and used automated tools to fill the feeds with spammy engagement-bait images. Using instructions and tools he bought online, 404 Media’s Jason Koebler set up his own AI slop shop. Here’s how it works:

“First, a spammer makes a free Vercel account. Then they make a WordPress account and spam page, download a snippet of code from GitHub that has been written by other spammers, modify it (there are guides for this on GumRoad and Fiverr; I bought guides on both), and install it on their Vercel deployment. Then they connect Vercel to the WordPress website using a specific WordPress plugin that effectively sets up a redirect between Vercel and the WordPress site. Then the person logs into their Facebook page and upgrades their account to an ad account, which is an instant process. Then the person goes through the process of buying and setting up an ad on Facebook that links to the Vercel mirror of the page they want to link to. This entire Vercel step exists to avoid Facebook’s automated spam detection, because it ultimately masks the link of the website that is being linked to. Within the ad manager, they upload an image that has been modified to look like a Facebook gallery post (there are tools for this), and then they delete the headline and description that is autogenerated by Facebook to make it look like the ad is just an image and not a link. Then, they generate a preview for their ad. This creates a preview link, which can then be pushed to their phone (it has to be a separate device), then shared to their own personal Facebook page as a ‘test.’ From there, the link can be grabbed and shared to a page as a normal post, even though it is not a published ad, meaning that, to do this trick, you do not actually have to spend any money. This process can then be repeated over and over again to effectively rick people into clicking on photo albums that are not actually photo albums. FewFeed has recently created a feature that allows you to directly create these fake photo album-type images.”

Profiting from AI slop is a technical process and Meta’s content moderation teams can’t keep up with the flood. Worse, they may not want to stop all grist-mill. Meta explicitly allows AI-generated images and the only metrics it cares about are the ones that improve its bottom line. The more people sharing posts and engaging with them, the better off Meta is. The slop makes everyone money.

“We encourage creators to use AI tools to produce high-quality content that meets all our Community Standards, and we take action against those who attempt to drive traffic using inauthentic engagement whether they use AI or not,” a Meta spokesperson told 404 Media. “We know bad actors adapt their tactics to evade our rules, so we are always working to improve our detection and enforcement.”

According to the people making the slop, the money is life-changing. Buried in the middle of the 404 Media story is a telling exchange. In one podcast, a spammer showed off his earnings tab. On the list was a picture of a train made of leaves. The spammer had earned $431 from the image’s engagement. “People don’t even make this much money in a month,” the interview said.