One in Five Videos Recommended to New YouTube Users Is AI-Generated for View Inflation

New YouTube users are highly likely to encounter a mass of low-quality, AI-generated videos. According to a Kapwing study, over 20% of the content the platform recommends to new accounts is so-called "AI slop."

Analysts manually examined the top 100 most popular channels in each country and found that 278 of these channels publish exclusively "AI slop." Collectively, such channels have amassed over 63 billion views and 221 million subscribers, with their estimated annual revenue reaching approximately $117 million. By creating a "clean" account, the researchers found that 104 of the first 500 YouTube Shorts suggested were AI-generated. If other low-quality content created solely for earning views is included, this share reaches one-third of all initial recommendations.

The phenomenon is global, but it is especially large-scale in some countries. Spain leads in the number of subscribers to such channels—20.22 million people, which is nearly half the country's population. In South Korea, 11 trending AI channels have garnered a record 8.45 billion views. The most-viewed channel in the sample is India's Bandar Apna Dost (2.4 billion views), where a talking monkey and a Hulk-like character battle demons. Kapwing estimates its annual revenue could reach $4.25 million. Meanwhile, the channel Imperio de jesus, offering quizzes featuring Jesus and Satan, has gathered 5.87 million subscribers.

In response to a request for comment, a YouTube spokesperson emphasized that generative AI is merely a tool and can be used to create both high-quality and low-quality content. The company stated it aims to promote quality content regardless of how it's created and removes videos that violate platform policies.

However, experts point to deeper problems. Researchers speak of an "illusory truth effect," where frequent exposure to low-quality or outright false AI-generated content can influence audience perception. The ease of creating convincing videos using neural networks opens opportunities for manipulation. So, despite the platform's efforts, recommendation algorithms still actively promote such material, especially to new audiences whose preferences are not yet formed.