Over 20% of videos shown to new YouTube users are AI slop, study finds

More than 20% of videos recommended to brand-new YouTube users are low-quality, AI-generated clips known as “AI slop,” according to a study released last month. Researchers at Kapwing analyzed the first 500 Shorts shown to a newly created account and identified 104 videos generated primarily by artificial intelligence. The same analysis found another 33% of recommendations consisted of similarly repetitive “brainrot” content.

Kapwing conducted the analysis using a new YouTube account with no viewing history, subscriptions, or user interactions, isolating the platform’s default recommendation behavior. The researchers examined Shorts because the format autoplays continuously and dominates early user exposure, allowing high-volume content to surface rapidly without active search.

The study defined AI slop as videos produced at scale using artificial intelligence with minimal human involvement, often repackaged across multiple channels with only slight visual or narrative variation. Brainrot content, tracked separately, was characterized by repetitive structures and fast pacing designed to maximize watch time rather than originality.

Kapwing found that South Korea accounts for one of the largest concentrations of AI slop viewership, with trending channels in the category amassing a combined 8.45 billion views. Many of these channels publish dozens of near-identical Shorts daily, increasing their likelihood of repeated algorithmic recommendation.

The most-viewed AI slop channel identified in the study was India’s Bandar Apna Dost, which recorded 2.07 billion views at the time of analysis. Using standard advertising revenue estimates, Kapwing projected the channel’s annual earnings at approximately $4,251,500, demonstrating how automated, low-effort content can generate significant revenue at scale despite platform restrictions on repetitive or inauthentic videos.

YouTube CEO Neal Mohan has defended the company’s expanding use of generative AI, proclaiming in a Sept. interview with Wired that the technology can “unlock creativity” for users who lack traditional editing or production skills. Mohan has framed AI tools as a way to broaden participation on the platform rather than replace human creators, even as automated content increasingly fills recommendation feeds.

Mohan has also acknowledged concerns about AI-generated imitation, particularly videos that copy the likeness or voice of real creators. He said YouTube is building tools that give creators “a choice as to whether that should come down,” while continuing to rely on automated systems to enforce content policies at scale as upload volume grows.