A surge in low-quality, AI-generated videos is increasingly dominating YouTube Shorts, with a new study revealing that more than half of short-video recommendations shown to new users fall into the category of so-called AI “slop” or brainrot content.
The findings come from research conducted by video-editing platform Kapwing and were first reported by The Guardian. The study highlights growing concerns about the quality of content promoted by social media algorithms.
Kapwing analysed the first 500 videos recommended to a newly created YouTube Shorts account with no viewing history or user interaction. The results showed that 104 videos, or 21 percent, were clearly AI-generated. A further 165 videos, representing 33 percent, were classified as brainrot — nonsensical, low-effort clips with little informational or creative value.
Combined, AI-generated and brainrot videos accounted for 54 percent of all recommendations presented to new users, raising questions about YouTube’s content moderation and recommendation systems.
Kapwing defines brainrot content as “nonsensical, low-quality video content that creates the effect of corroding the viewer’s mental or intellectual state.” The study noted that such content is frequently AI-generated, making it increasingly difficult to distinguish between automated output and human-made low-effort videos.
The prevalence of AI slop varies significantly by region. Spain emerged as the country with the highest combined subscriber count for AI slop channels, totalling more than 20 million subscribers, despite having fewer such channels among its top 100 creators. The United States ranked third globally, with AI slop channels attracting over 14 million subscribers and nine such channels appearing among its top 100 most popular accounts.
Researchers note that YouTube is not alone in facing this issue. Similar patterns of AI-generated, low-quality content have been observed across major social media platforms, including viral fake animal footage and clearly fabricated videos designed to drive engagement.
Kapwing’s study suggests that the spread of AI slop is unlikely to slow down in the near future, as easy-to-use AI tools and engagement-driven algorithms continue to reward frequent uploads over content quality.
