You will be amazed by the sheer volume of AI-generated garbage and mindless videos on YouTube.

      A recent study conducted by Kapwing reveals that a significant amount of videos on YouTube consists of what is described as brainrot content and ‘AI slop’: a term that refers to low-quality, AI-generated material primarily created to gain views instead of providing genuine value.

      To investigate this, Kapwing's researchers set up a new YouTube account and monitored the first 500 videos the platform recommended. They found that 104 of these videos, approximately 21%, were identified as AI slop, while 165 videos, around 33%, were grouped into a broader "brainrot" category.

      Brainrot encompasses repetitive, bizarre, or hypnotic videos that are easy to consume but lack depth. The results imply that a substantial portion of content recommended to new users is automated rather than produced by actual creators.

      The extent of AI slop on YouTube is significant, extending beyond just a handful of odd suggestions. Kapwing also examined trending YouTube channels in various countries and discovered 278 channels composed entirely of AI slop, present within the global top 100 rankings.

      These channels are not insignificant; collectively, they have garnered billions of views and millions of subscribers, resulting in tens of millions of dollars in estimated annual advertising revenue. Certain regions are particularly notable; for example, AI slop channels in Spain boast over 20 million combined subscribers, outpacing totals in the United States or Brazil.

      In South Korea, these slop channels have generated over 8.45 billion views, while India's leading AI slop channel has alone exceeded 2 billion views. These statistics indicate that AI slop is not restricted to a single market but is expanding worldwide.

      The rapid expansion of this content can be attributed less to individual creators and more to the incentives embedded within recommendation algorithms. Videos generated by AI are inexpensive to create, can be uploaded on a large scale, and are frequently optimized to evoke curiosity or encourage endless scrolling.

      New users are particularly susceptible because the algorithm lacks a viewing history to guide its recommendations. For YouTube, these findings pose challenging questions. If 20% of initial recommendations consist of AI slop videos, this could alter the way users perceive YouTube before they even discover creators they genuinely wish to watch.

      While YouTube has introduced tools aimed at addressing deepfakes, I would like to see the platform implement improved controls to limit AI slop, similar to the measures TikTok has already adopted. A report from Amazon Web Services (AWS) suggests that 57% of the internet may already comprise AI sludge.

      This leads DuckDuckGo to provide tools for filtering out low-quality AI content, whereas some tools like Slop Evader take it a step further by restoring the web to its appearance before generative AI became prevalent.

      As AI tools facilitate the inundation of platforms with synthetic media, the challenge will be determining whether engagement alone should dictate what new viewers encounter first.