The study found that more than 20% of videos suggested to first-time users were low-quality, AI-generated clips designed primarily to garner views.
Super Cat League features human-like cats in bizarre scenes. Photo: SuperCatLeague/YouTube
“>
Super Cat League features human-like cats in bizarre scenes. Photo: SuperCatLeague/YouTube
A growing portion of the content recommended to new YouTube users is what researchers call “AI slop,” according to a recent study examining the platform's recommendation system.
The study found that more than 20% of videos suggested to first-time users were low-quality, AI-generated clips designed primarily to garner views. According to the Guardian, researchers estimate that this rapidly expanding sector generates around $117 million in revenue annually across major social media platforms.
The analysis found that 278 of the world's most popular YouTube channels are made up entirely of this type of content. Combined, these channels have over 63 billion views and over 221 million subscribers.
Researchers say these channels have large audiences in multiple regions, including Spain, Egypt, the United States, and Brazil. Many of the videos seem to target children through bright animation, while others rely on surreal or unscripted scenes featuring anthropomorphic animals or superheroes.
Some creators have been found to be using artificial intelligence to create “creepy” depictions of natural disasters, such as the massive floods described in the study, often paired with calming background music. Researchers said these videos are designed to maximize watch time and ad revenue.
The study notes that the industry is largely driven by creators based in middle-income countries such as India, Nigeria, and Kenya, and revenue from YouTube can exceed local median wages.
YouTube said its recommendation system aims to connect users with high-quality content, whether created by humans or generated using AI tools. But critics argue that large platforms act as “A/B testing machines,” prioritizing content that performs well in their algorithms rather than evaluating originality or craftsmanship.
The researchers likened the phenomenon to “digital junk food,” describing it as mass-produced content created using cheap, automated methods to trigger algorithmic engagement while offering limited artistic value.
