Over 21% of YouTube Shorts is AI slop, with creators earning millions farming views

1 in 5 YouTube Shorts is AI slop now - and Americans are eating it up

A new study shows AI slop makes up more than 20% of YouTube.Brainrot, or poor quality content, makes up 33%.There's no incentive for creators to try to lessen AI slop. If you feel like you're seeing a lot more AI-generated slop on YouTube, you're right. According to a study from video-editing service Kapwing, more than 20% of the Shorts content on YouTube is AI slop, or "careless, low-quality content" made with AI and designed solely to farm views and subscriptions. The study showed that 33% is what's called brainrot, or "compulsive, nonsensical, low-quality video content" (that's often generated with AI). Also: How I download YouTube videos for free - 2 easy and reliable methods to do it To conduct the study, Kapwing took the top 100 trending YouTube channels in every country and noted which ones were AI-generated. It then used a social media analytics tool to track the overall views, subscriber count, and estimated yearly revenue for those channels to determine their popularity. It also created a new YouTube account to see how much AI slop was offered up among the first 500 Shorts. In theory, an algorithm-free account should be a good overview of YouTube content at large, since there aren't any personalized recommendations yet. In the scenario with a totally new account, it took 16 Shorts before YouTube served one that was AI slop. From there, though, it was a fairly steady stream, as 104 of the first 500 videos, or 21%, were low-quality, AI-generated. Another 165 of the first 500 videos, or 33%, fell into the brainrot category. On a global scale, Spain was the largest consumer of AI slop, with more than 20 million people subscribing to trending AI channels. Egypt took second place, with the US in third. The most popular AI channel, Kapwing found, is Bandar Apna Dost, an Indian channel with 2.07 billion views. It brings in an estimated $4.25 million a year. Also: Stop accidentally sharing AI videos - 6 ways to tell real from fake before it's too late Of course, YouTube only serves up what it sees as popular or engaging, so people must be watching this style of AI video. When you consider that for creators, AI generators mean very low-cost or even no-cost content, it's not hard to see why there's been a sudden rise. And if YouTube reinforces that by pushing the videos and rewarding creators, the rise will only continue. Many platforms, like Meta (Facebook, Instagram, and Threads) and Pinterest, are taking steps to label content, and some AIs, like Gemini, can now detect whether or not a video was made by AI. As long as users keep consuming it, though, there's no incentive for creators or platforms to try to lessen AI slop.

Over 21% of YouTube Is Now AI Slop, Says Report

As tools like OpenAI's Sora and Google's Veo are making AI video creation more accessible than ever, over 21% of YouTube is now what's called "AI slop," according to a new report by video editing firm Kapwing. The company simulated the experience of an "untainted" YouTube Shorts algorithm by establishing a new YouTube account and noting the prevalence of AI slop or "brainrot" videos -- low-quality, trivial online content -- among the first 500 videos that showed up in the feed. Though the test accounts were spared either of these for the first 16 videos in the feed, in total, 104 (21%) of the first 500 videos were AI-generated. South Korea, which has some of the highest rates of social media and smartphone use globally, is leading the way in terms of slop consumption, viewing its top AI slop channels roughly 8.25 billion times. The East Asian nation's top slop channel, Three Minutes Wisdom, which features photorealistic-style footage of wild animals being defeated by cute household pets, has racked up 2.02 billion views by itself. Pakistan came in second place in the rankings, with its top slop channels racking up 5.34 billion views. The United States ranked third in terms of slop consumption, with the top channels collecting 3.39 billion views. The top channel in the US, the Spanish-language Cuentos Facientes, has collected 1.28 billion views and earned an estimated $2.66 million for its creators. But while YouTube may be more slop-infested than ever, it's not the only part of the web being overrun by AI-generated content as of late. Written content is now more prominently AI-generated than ever, with the percentage of AI articles on the web now slightly above 50% as of May 2025. Meanwhile, even professional workplaces aren't immune from AI-generated content. Four in ten employees in the US have received "workslop" material over the past year consisting of AI-generated work content that "masquerades as good work but lacks the substance to meaningfully advance a given task," with sectors like IT and consulting being particularly hard hit. Some YouTube competitors, such as TikTok, have taken steps to stem the tide of AI slop on their platforms. TikTok rolled out features that can enable users to keep AI slop out of their feeds last month to help users better identify AI-generated content.

More than 20% of YouTube's feed is now "AI slop," report finds

Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust. The takeaway: Not only are more than half the articles on the web created by AI, but over 21% of YouTube videos being show to new users is "AI Slop." That's according to a new report, which also found that the US is in third place when it comes to consumption of these low-quality generated videos. Video editing firm Kapwing highlights AI slop's definition as careless, low-quality content generated by computer applications and used to farm views and subscriptions or sway political opinion. But for all the pushback against AI slop and brainrot - low-quality, trivial online content -- their proliferation on YouTube is hard to avoid. To get an idea of how much of YouTube consists of AI-generated videos, Kapwing simulated the experience of an untainted YouTube Shorts algorithm by establishing a new YouTube account. It then noted the occurrence of AI slop among the first 500 videos in the feed. In total, 104 (21%) of the first 500 videos were AI-generated, while 165 (33%) were classed as brainrot. Kapwing also looked at global consumption of AI slop. Based on views alone, South Korea leads the way - its 11 trending AI channels have a combined figure of over 8.45 billion views. Pakistan is second with 5.34 billion views, while the US is third with 3.39 billion. It's estimated that the Three Minutes Wisdom channel, which accounts for almost a quarter of the country's view count, earns around $4 million annually from ad income. The US is also third when it comes to the most subscribers to these channels. The 14.4 million subscribers in the country put it just beneath second-place Egypt (17.9 million) and leader Spain (20.22 million). The world's most-subscribed AI slop channel is a US-based Spanish-language channel called Cuentos Facinantes [sic], which has 5.95 million subs. Its Dragon Ball-themed videos have racked up 1.28 billion views. India's Bandar Apna Dost has the most views of any AI channel: 2.07 billion. It's also the highest earner, with an estimated $4.25 million per year. It's not just YouTube, of course. Social media, online articles, images - huge swathes of what was once almost entirely the work of humans are now generated by AI. It's no wonder that "slop" is Merriam-Webster's word of the year.

Study reveals just how much AI slop is on YouTube

If it feels like there's a lot of AI slop on YouTube, that's because there's a lot of AI slop on YouTube. New research from video-editing company Kapwing, reported by the Guardian found that more than one in every five videos that the YouTube Shorts algorithm shows new users is low-quality, AI-generated content. One of the most interesting parts of the Kapwing study is that of the first 500 YouTube Shorts videos in a brand-new, untouched YouTube Shorts algorithm, 104 were AI-generated and 165 were brainrot -- a whopping 21 percent and 33 percent, respectfully. Of course, the love of AI slop differs depending on the country. Kapwing found that AI slop channels in Spain have a combined 20.22 million subscribers, more than any other country, but has fewer AI slop channels among its top 100 channels than other countries. The U.S. has nine channels among its top 100 channels, and the third-most slop subscribers at 14.47 million. YouTube isn't the only social media beast whose content is falling to the depths of AI slop despair, but the Kapwing study makes it clear that AI slop isn't going anywhere. As Mashable's Tim Marcin reported earlier this month, AI slop is taking over our feeds, from fake animals on surveillance tapes to heavy machinery cleaning barnacles off whales.

From shrimp Jesus to erotic tractors: how viral AI slop took over the internet

Flood of unreality is an endpoint of algorithm-driven internet and product of an economy dependent on a few top tech firms In the algorithm-driven economy of 2025, one man's shrimp Jesus is another man's side hustle. AI slop - the low-quality, surreal content flooding social media platforms, designed to farm views - is a phenomenon, some would say the phenomenon of the 2024 and 2025 internet. Merriam-Webster's word of the year this year is "slop", referring exclusively to the internet variety. It came about shortly after the advent of popular large language models, such as ChatGPT and Dall-E, which democratised content creation and enabled vast swathes of internet denizens to create images and videos that resembled - to varying degrees - the creations of professionals. In 2024, it began to achieve peak cultural moments. Notable among these was shrimp Jesus, a viral trend in which Facebook was briefly flooded with AI-generated images of the deity fused with crustaceans. Shrimp Jesus was quickly followed by hallmarks of the AI slop genre: videos of old women claiming to celebrate their 122nd birthday, and mini soap operas about the dramatic lives of cats. In 2025, the flood continued, growing more uncanny and more explicitly copyright-violating. This spring saw the advent of Ghiblification - that is, a trend in which users from Nayib Bukele to the White House rendered images, including of deportations, in the style of Hayao Miyazaki's Studio Ghibli. This particular moment was enabled by OpenAI's release of an image generator powered by GPT-4o; Sam Altman, OpenAI's chief executive, jumped on the trend by Ghiblifying his X profile and writing the rather remarkable post: Miyazaki, the chief architect of Studio Ghibli's distinctive, hand-drawn animation style, has elsewhere said, on the subject of artificial intelligence: "I would never wish to incorporate this technology into my work at all. I strongly feel this is an insult to life itself." Other AI slop moments followed: a spate of videos of AI-generated obese people participating in the Olympics, pressure cookers exploding, more cats. Ibrahim Traoré, the leader of the military junta in Burkina Faso, became the centrepiece of an AI slop cult featuring videos of Justin Bieber singing on the streets of Ouagadougou. In some ways, AI slop has improved. Gone - mostly - are the days of six-fingered hands and missing limbs that characterised the output of early image generators. In some ways, though, AI slop has hardly changed at all. It is still uncanny and contextless, still aimed directly at the amygdala, still chasing virality by virtue of having the lowest barriers to entry imaginable: no plot, no exposition, surreal imagery and cats, cats, cats. Describing this flood of unreality as solely a technological phenomenon misses one of the main drivers of AI slop. In one view it is the endpoint of an algorithmically determined internet optimised for engagement and turbocharged with new, powerful tools. It won't change as long as the platforms and their algorithms don't. But it is also the product of an underlying global economy - one that is everywhere, increasingly dependent on a few powerful technology companies and a few powerful platforms, one which appears to offer dwindling returns for real work, but lavish fortunes for the lucky, viral few. AI slop creator is, after all, a profession. They come from everywhere - from the US to India to Kenya to Ukraine. In some ways, it can be argued that AI tools have enabled a strange globalisation of content, says Arsenii Alenichev, who studies the production of images in global health. Earlier this year, Alenichev noticed a flood of "AI poverty porn" on major stock photo sites. Many of the creators of the images, he said, appeared to have eastern European usernames. "I wouldn't be surprised if these are just artists that are trying to generate extreme images of everything, hoping that someone would buy them," he said. Making it in AI slop isn't easy. Oleksandr, an AI YouTube creator based in Chernivtsi, Ukraine, estimates that only the top 5% of creators ever monetise a video, and only 1% make a living from it. Oleksandr began his business in 2024, after retiring from being a professional volleyball player. He was deep in debt, he said, and at a low point in his life: his girlfriend had left him, his parents were living in occupied Mariupol. He started to join Telegram channels and watch YouTube videos on how to make money from YouTube. His first efforts were music channels, playing AI-generated music over images of sexy AI girls. He had seven: retrowave, rock, jazz and more. At first he put a great deal of effort into each video, he said, but he realised quickly that that didn't matter on YouTube. "It was a conveyor belt, with fairly low quality." His videos got attention, standing out from hundreds of other similar channels, even to the point where a Japanese film-maker contacted him to license one of his pieces for a short movie. Then he expanded. At the high point of his business, he had a team of 15 people operating 930 channels, 270 of which he successfully monetised. They cleared up to $20,000 (£15,000) a month at one point, although YouTube often blocked or took down his channels - including the sexy AI girls - for unclear reasons. His content evolved. One fruitful niche he found was life stories - long anecdotes written by ChatGPT or Gemini, overlaid with visuals, which were extremely popular: "Grandparents listen to it before bed, or while walking in the park." Another niche, he said, was videos on "vulgar adult themes" - such as erotic tractors - which were in great demand, but bordering on what YouTube allows. These channels were riskier to produce, but at times were easier to monetise, because they had less competition. With "erotic ones it's easier, because they are blocked more often, so not many people want to bother and periodically recreate channels", he said. "I saw the opportunity, and other people saw the difficulty." Now, the work has shrunk somewhat: YouTube has become more aggressive with its takedowns of content, meaning he has to recreate channels. He and his team now take in closer to $3,000. But Oleksandr credits the platform - and the videos he watched - with changing his life, allowing him to resolve his debts and build a career he (somewhat) loves. It is not a site for artistic aspirations, though. A fair amount of the work involves adding nearly naked AI women to videos of tractors. "To make money here, you need to spend as little as possible," he said. "YouTube is basically just clickbait and sexualisation, no matter how morally sad it is. Such is the world and the consumer." A YouTube spokesperson said: "Generative AI is a tool, and like any tool it can be used to make both high- and low-quality content. We remain focused on connecting our users with high-quality content, regardless of how it was made. All content uploaded to YouTube must comply with our community guidelines, and if we find that content violates a policy, we remove it."

The amount of AI slop and brainrot videos on YouTube is shocking

A new study by Kapwing has says that a huge volume of videos circulating on YouTube are brainrot content and 'AI slop': a catch-all term for low-quality, AI-generated content designed to farm views rather than offer real value. Kapwing's researchers tested this by creating a new YouTube account and tracking the first 500 videos recommended by the platform. Out of those, 104 videos, roughly 21%, were classified as AI slop, while 165 videos, about 33%, fell into a broader "brainrot" category. Recommended Videos Brainrot includes repetitive, bizarre, or hypnotic clips that are easy to watch but lack substance. Together, the findings suggest that a significant chunk of what new users see is automated content rather than work made by human creators. How much AI slop YouTube is actually serving The scale of such content goes far beyond a few strange recommendations. Kapwing also analyzed trending YouTube channels across multiple countries and found 278 channels made entirely of AI slop, spread across the global top 100 rankings. These channels are not small. Collectively, they have accumulated billions of views and millions of subscribers, translating into tens of millions of dollars in estimated annual ad revenue. Some regions stand out in particular. In Spain, AI slop channels have more than 20 million combined subscribers, higher than totals seen in the United States or Brazil. South Korea's slop channels have generated over 8.45 billion total views, while India's largest AI slop channel alone has surpassed 2 billion views. These rankings show that AI slop is not confined to one market, but is spreading globally. Why is it spreading so fast? The problem is less about individual creators and more about incentives baked into recommendation algorithms. AI-generated videos are cheap to produce, can be uploaded at a massive scale, and are often optimized to trigger curiosity or endless scrolling. New users are especially vulnerable because the algorithm has no viewing history to guide recommendations. For YouTube, the findings raise uncomfortable questions. If a fifth of early recommendations are AI slop videos, it could reshape how users experience YouTube before they ever find creators they actually want to watch. While YouTube has rolled out tools designed to curb deepfakes, I would like to see the platform offering better controls to limit AI-slop, much like TikTok already does. A report by Amazon Web Services (AWS) researchers claims that 57% of the internet may already be AI sludge. This is why DuckDuckGo offers tools to filter low-quality AI content, while some tools like Slop Evader go even further by stripping the web back to how it looked before generative AI took over. As AI tools make it easier to flood platforms with synthetic media, the challenge will be deciding whether engagement alone should keep driving what new viewers see first.

More than 20% of videos shown to new YouTube users are 'AI slop', study finds

Low-quality AI-generated content is now saturating social media - and generating about $117m a year, data shows More than 20% of the videos that YouTube's algorithm shows to new users are "AI slop" - low-quality AI-generated content designed to farm views, research has found. The video-editing company Kapwing surveyed 15,000 of the world's most popular YouTube channels - the top 100 in every country - and found that 278 of them contain only AI slop. Together, these AI slop channels have amassed more than 63bn views and 221 million subscribers, generating about $117m (£90m) in revenue each year, according to estimates. The researchers also made a new YouTube account and found that 104 of the first 500 videos recommended to its feed were AI slop. One-third of the 500 videos were "brainrot", a category that includes AI slop and other low-quality content made to monetise attention. The findings are a snapshot of a rapidly expanding industry that is saturating big social media platforms - from X to Meta to YouTube - and defining a new era of content: decontextualised, addictive and international. A Guardian analysis this year found that nearly 10% of YouTube's fastest-growing channels were AI slop, racking up millions of views despite the platform's efforts to curb "inauthentic content". The channels found by Kapwing are globally distributed and globally watched. They have millions of subscribers: in Spain, 20 million people, or nearly half the country's population, follow the trending AI channels. AI channels have 18 million followers in Egypt, 14.5 million in the US, and 13.5 million in Brazil. Bandar Apna Dost, the most-viewed channel in the study, is based in India and now has 2.4bn views. It features the adventures of an anthropomorphic rhesus monkey and a muscular character modelled off the Incredible Hulk who fights demons and travels on a helicopter made of tomatoes. Kapwing estimated that the channel could make as much as $4.25m. Its owner did not respond to a query from the Guardian. Rohini Lakshané, a researcher on technology and digital rights, said Bandar Apna Dost's popularity most likely stems from its absurdity, its hyper-masculine tropes and the fact that it lacks a plot, which makes it accessible to new viewers. Pouty Frenchie, based in Singapore, has 2bn views and appears to target children. It chronicles the adventures of a French bulldog - driving to a candy forest, eating crystal sushi - many of them set to a soundtrack of children's laughter. Kapwing estimates it makes nearly $4m a year. Cuentos Facinantes, based in the US, also appears to target children with cartoon storylines, and has 6.65 million subscribers - making it the most-subscribed channel in the study. Meanwhile, The AI World, based in Pakistan, contains AI-generated shorts of catastrophic flooding in Pakistan, with titles like Poor People, Poor Family, and Flood Kitchen. Many of these videos are set to a soundtrack called Relaxing Rain, Thunder & Lightning Ambience for Sleep. The channel itself has 1.3bn views. It's hard to say exactly how significant these channels are compared with the vast sea of content already on YouTube. The platform does not release information on how many views it has yearly, or how many of these are from AI content. But behind these uncanny scenes of candy forests and disasters is a semi-structured, growing industry of people trying to find new ways to monetise the world's most powerful platforms using AI tools. "There are these big swathes of people on Telegram, WhatsApp, Discord and message boards exchanging tips and ideas [and] selling courses about how to sort of make slop that will be engaging enough to earn money," said Max Read, a journalist who has written extensively on AI slop. "They have what they call niches. One that I noticed recently is AI videos of people's pressure cookers exploding on the stove." While creators of AI slop are everywhere, Read said that many come from English-speaking countries with relatively strong internet connectivity, where the median wage is less than the amount they can make on YouTube. "It's mostly sort of middle-income countries like Ukraine, lots and lots of people in India, Kenya, Nigeria, a fair number in Brazil. You see Vietnam, too. Places with relative freedom online to access social media sites," he said. It's not always easy to be an AI slop creator. For one thing, creator programmes on YouTube and Meta aren't always transparent about who they pay for content, and how much, said Read. For another, the AI slop ecosystem is full of scammers: people selling tips and courses on how to make viral content - who often make more money than the AI slop producers themselves. But, at least for some, it's a living. And while new, attention-grabbing ideas - such as exploding pressure cookers - constantly emerge, when it comes to AI slop, human creativity matters far less than the algorithms that distribute the content on Meta and YouTube. "These websites are huge A/B testing machines just by their nature," said Read. "Almost anything that you can think of, you could already find on Facebook. So the question is, how do you find the things that are kind of doing well, and then how do you scale that? How do you make 10 of them?" A YouTube spokesperson said: "Generative AI is a tool, and like any tool it can be used to make both high- and low-quality content. We remain focused on connecting our users with high-quality content, regardless of how it was made. All content uploaded to YouTube must comply with our community guidelines, and if we find that content violates a policy, we remove it."

AI Slop Videos Make Up 33% of YouTube Feed, Says Study

Low-quality AI-generated videos now make up a significant share of what users see on YouTube, raising concerns about platform quality, monetisation, and misinformation, according to new research by Kapwing. The study estimates that 21-33% of YouTube's content feed may consist of what it describes as "AI slop" or "brainrot" videos. These are cheaply produced, repetitive videos generated using AI tools, designed mainly to attract views rather than provide meaningful content. Kapwing's analysis also suggests that some of these channels may be earning millions of dollars annually, despite minimal human effort or creative input. Kapwing examined the top 100 trending YouTube channels in each country, identifying those that primarily publish AI-generated content. The company then used Social Blade data to assess views, subscribers, and estimated earnings. To understand what new users see, researchers also created a fresh YouTube account and reviewed the first 500 Shorts shown in the feed. Among the most striking findings: The research shows that AI slop does not need large numbers of YouTube channels to dominate attention. In Spain, for example, only eight AI slop channels appeared among the top 100 trending channels, yet they attracted more subscribers than countries with far more such channels. In South Korea, a single channel, Three Minutes Wisdom, accounted for nearly a quarter of the country's total AI slop views, with 2.02 billion views. Kapwing estimates that this channel alone could earn over $4 million per year from ads. Many of these videos follow simple, repetitive formats, such as animals in dramatic situations, quiz-style religious content, or fictional characters in endless variations of the same scenario, making them cheap and fast to produce at scale. Kapwing's findings align with how YouTube Shorts are designed and distributed. Shorts rely heavily on algorithmic recommendations rather than subscriber-based feeds, increasing exposure for high-volume publishers regardless of prior audience size. Short-form video platforms reward frequent uploads, quick watch times, and continuous scrolling, which makes the format particularly well-suited to AI-generated content that creators can produce and publish rapidly with minimal variation. India is home to the most-viewed AI slop channel identified in the study. Bandar Apna Dost uploads hundreds of near-identical videos featuring a digitally generated monkey placed in emotional or dramatic situations. The channel also maintains a presence on Instagram and Facebook, where it reposts similar content. Kapwing's revenue estimates indicate that high-volume AI-generated content from India can compete directly with professionally produced videos in terms of reach and earnings. YouTube enforces its monetisation rules under the YouTube Partner Program across all regions where the program is available. YouTube has expanded these monetisation rules and enhanced content detection systems to apply across all markets participating in the Partner Program, ensuring consistent enforcement of policy standards regardless of a creator's location. The study does not suggest that India produces more AI-generated content by design but highlights how global recommendation systems and ad models reward scale over production method. YouTube currently allows the use of generative AI tools in video creation and does not prohibit monetisation of AI-generated videos by default. The platform requires creators to disclose altered or synthetic media only in limited cases, such as when content could mislead viewers about real-world events or individuals. YouTube has not publicly defined a threshold at which AI-generated or repetitive content becomes spam or ineligible for monetisation. It continues to recommend and monetise videos that comply with its community guidelines and advertiser-friendly content rules. YouTube CEO Neal Mohan has defended generative AI in video creation, saying the level of AI involvement does not determine content quality. In an interview with Wired, he said: "The genius is going to lie in whether you did it in a way that was profoundly original or creative. Just because the content is 75% AI-generated doesn't make it any better or worse than a video that's 5% AI-generated. What's important is that it was done by a human being." Advertiser concerns highlighted by the study relate primarily to ad placement and adjacency, rather than direct policy violations. Industry discussions increasingly distinguish between harmful content and low-quality or repetitive content, with the latter raising questions about brand perception rather than safety. While YouTube has not announced advertiser-specific restrictions on AI-generated videos, advertisers can already use existing brand safety tools to limit placements by content category, format, or channel. Kapwing's test of a new YouTube account showed how quickly AI slop appears in the Shorts feed. Of the first 500 videos shown: The study notes that it remains unclear whether algorithms deliberately promote such content or whether its visibility results from the overwhelming volume of uploads. External reporting supports the trend. An analysis cited by The Guardian found that nearly one in 10 of the world's fastest-growing YouTube channels publish only AI-generated content. The Kapwing study does not assess whether AI-generated content directly displaces human creators in recommendations, but documents that a small number of AI-driven channels can account for a disproportionate share of views and impressions across multiple markets. Researchers and media analysts warn that the spread of AI slop goes beyond entertainment quality. Repetitive exposure to fabricated or misleading visuals can reinforce false beliefs, a phenomenon known as the "illusory truth effect. AI tools also lower the cost of producing politically or ideologically motivated content at scale. Artist and researcher Eryk Salvaggio has argued that an excess of information eventually turns into noise and increases public dependence on algorithms to decide what people see and trust. The Kapwing report concludes that while AI-generated video is not inherently harmful, the unchecked growth of low-quality content may reshape online media in ways that prioritise volume and engagement over credibility, creativity, and trust. Data referenced in the study is accurate as of October 2025.