More than one in five videos recommended for new YouTube accounts are low-quality clips generated by AI, aptly known as “AI slop,” according to new research from Kapwing. These videos, which have also taken social media by storm, are primarily created to get people's attention and generate advertising revenue, but they're not as harmless as they sound. According to Kapwing's research, this tilt is completely reshaping the YouTube ecosystem as we know it.
In its research, Kapwing analyzed the performance of global channels and the experience of newly created YouTube accounts. Researchers found that 104 of the first 500 videos served to a new account were generated by AI. That's over 20%. This finding suggests that YouTube's recommendation system isn't doing this by chance. Instead, you habitually and intentionally expose new users to it. AI lag isn't the only issue. Kapwing also found that a third of the first 500 videos were obsessive, low-quality, “brain rot” videos, whether generated by AI or not.
Other troubling trends revealed by the data
Kapwing's findings also found impacts beyond individual feeds. When we looked at the top 100 trending YouTube channels in each country, we found that hundreds of the top channels were almost entirely dedicated to AI-generated videos. Together, these channels have garnered tens of billions of views and hundreds of millions of subscribers worldwide. This equates to an estimated annual revenue of $117 million. And who knows how much money could be made from fake AI scam ads featuring celebrities on YouTube?
The investigation found that these channels mainly upload surreal or absurd short animations with no clear plot, or bright looping animations aimed at children. Some of the largest audiences for these channels include Spain, South Korea, Egypt, Brazil, and India. In fact, India-based channel Bandar Apna Dost was named as the most watched channel in the entire study. This one channel alone has over 2 billion views.
Despite the study's findings, YouTube insists it will not ban generative AI as long as channels disclose their use. However, some questions still remain. Why remain neutral when the site's recommendation system is amplifying such low-value content? What if the situation worsens enough to drive out top creators? As AI advances, YouTube will need to consider this. In the meantime, if you're tired of the AI slops and brains in your feed, it's helpful to know how to improve your recommendations.
