More than a fifth of the content circulating on YouTube now falls into what researchers describe as “AI slop,” according to a recent analysis by Kapwing. The term refers to mass-produced, low-effort videos generated with artificial intelligence tools and designed primarily to exploit recommendation algorithms rather than offer meaningful entertainment or information. As AI video software becomes easier to use and cheaper to deploy, the volume of this material is growing faster than most moderation systems can realistically track.
Kapwing’s findings were based on a controlled experiment that attempted to mirror how a new user might encounter content on YouTube Shorts. Researchers created a fresh account with no viewing history and observed the first 500 videos served by the algorithm. While the initial batch avoided obvious examples of AI-generated filler, the pattern shifted quickly. Out of those 500 videos, 104 were identified as AI-generated “brainrot,” accounting for just over 21 percent of the feed. The result suggests that algorithmic discovery, rather than user preference alone, is playing a central role in amplifying this kind of content.
The report also highlights significant regional differences in consumption. South Korea ranked first, with leading AI slop channels collectively attracting an estimated 8.25 billion views. One of the most watched examples, Three Minutes Wisdom, relies on AI-generated imagery depicting implausible scenarios involving animals and pets, a format that performs well in short-form feeds. Pakistan followed with 5.34 billion views across its top channels, while the United States placed third at 3.39 billion views. In the US market, a Spanish-language channel called Cuentos Facientes alone accounted for more than a billion views, translating into an estimated $2.66 million in ad revenue.
The rise of AI slop on YouTube mirrors a broader shift across the internet. By mid-2025, AI-written material reportedly made up slightly more than half of all newly published online articles. The same dynamic is showing up in professional environments, where automated text and presentations increasingly circulate as finished work. Surveys cited in the report indicate that roughly 40 percent of US employees have encountered internally shared AI-generated material that appears polished but contributes little substance, particularly in IT and consulting roles.
Advances in tools such as OpenAI’s Sora and Google’s Veo have lowered the barrier to producing convincing video at scale, making it harder for platforms to distinguish between creative experimentation and industrial content farming. Competing platforms are beginning to respond. TikTok recently introduced clearer labels and user controls aimed at reducing the visibility of AI-generated videos in personal feeds.
For YouTube, the challenge is less about banning AI outright and more about deciding how much algorithmic space this material should occupy. As generative tools continue to improve, the line between harmless novelty and feed pollution is likely to become even harder to draw, raising ongoing questions about quality, authenticity, and the incentives driving online video ecosystems.
