Over 20% of YouTube’s Recommendations to Newbies Are AI-Generated Video, Kapwing Study Reveals

Add us on Google

A new report from video-editing startup Kapwing has pulled back the curtain on YouTube’s algorithmic underbelly, revealing that more than 20% of videos pushed to fresh accounts are straight-up “AI slop” — that is, cheaply generated content engineered purely for clicks and ad dollars.

Kapwing’s team dug into 15,000 top YouTube channels (the leading 100 from each country) and identified 278 that are pumping out nothing but this AI-fueled filler. These channels have collectively snagged over 63 billion views and 221 million subscribers, translating to roughly $117 million in yearly revenue, based on industry estimates.

To test the waters firsthand, the researchers spun up a brand-new YouTube account and monitored the first 500 recommended videos. Of those, 104 were pure AI slop, while a full third qualified as “brainrot”, a broader bucket including AI junk and other attention-sucking low-quality fare.

This isn’t just a YouTube quirk; it’s a symptom of a booming shadow economy that’s infiltrating X, Meta, and beyond. We’re talking decontextualized, hyper-addictive content that’s going global at warp speed. Earlier this year, a Guardian dive found that nearly 10% of YouTube’s fastest-rising channels were AI slop operations, amassing millions of views despite Google’s crackdowns on “inauthentic” stuff.

Kapwing’s findings show these channels are a worldwide phenomenon. In Spain, trending AI accounts boast 20 million followers, which is almost half the population. Egypt clocks in at 18 million, the U.S. at 14.5 million, and Brazil at 13.5 million.

Take India’s Bandar Apna Dost, the study’s most-watched offender with 2.4 billion views. It stars an anthropomorphic monkey and a Hulk-like brute battling demons or cruising in a tomato helicopter. Kapwing pegs its potential earnings at up to $4.25 million. (We reached out to the owner; crickets so far.)

Rohini Lakshané, a tech and digital rights researcher, attributes the appeal to its sheer weirdness, macho vibes, and plotless format, perfect for dipping in anytime.

Then there’s Singapore-based Pouty Frenchie, with 2 billion views, seemingly aimed at kids: a French bulldog zipping to candy forests or munching crystal sushi, all backed by giggling soundtracks. Estimated haul? Nearly $4 million a year. U.S.-based Cuentos Facinantes, another kid-targeted cartoon channel, leads in subs with 6.65 million.

Pakistan’s The AI World takes a darker turn, churning out AI shorts of floods devastating poor families and kitchens, often scored to “Relaxing Rain, Thunder & Lightning Ambience for Sleep.” It’s pulled 1.3 billion views.

Quantifying the full scale is tough; YouTube doesn’t spill on total annual views or AI breakdowns. But beneath the bizarre visuals lies a semi-professional hustle: folks worldwide leveraging AI to game Big Tech’s algorithms for cash.

Journalist Max Read, who’s chronicled this trend, points to underground networks on Telegram, WhatsApp, Discord, and forums where creators swap tactics, hawk courses, and chase “niches” like exploding pressure cookers.

Read notes that many slop-makers hail from mid-income spots with solid internet but lower wages, think Ukraine, India, Kenya, Nigeria, Brazil, Vietnam, where YouTube payouts beat local gigs.

It’s not all smooth sailing: Payout programs from YouTube and Meta are opaque, and the space is rife with grifters peddling “viral secrets” who often out-earn the creators they’re “teaching.”

Still, for those who crack the code, it’s a viable income. Human ingenuity? Secondary. As Read puts it, these platforms are massive A/B testing labs: Spot what’s popping, then scale it to infinity.

A YouTube spokesperson said, “Generative AI is just a tool; it can create gems or garbage. Our priority is surfacing quality content, no matter the method. Everything on YouTube must follow our guidelines; violators get yanked.”