I didn't expect a spreadsheet to make my stomach turn. But that's what happened when I sat down with the numbers behind YouTube's newest, strangest income stream — one built almost entirely on content nobody asked for, made by nobody in particular, and watched by almost everybody.
We've all felt it. You open YouTube, meaning to watch one specific video, and somewhere between the second and fifth recommendation, you land on something that feels off. A dog with too many teeth. A monkey riding a tomato-shaped helicopter. A voice that's a little too smooth, saying nothing in particular. You scroll past it. Most people do.
But a small group of researchers didn't scroll past. They sat down, opened 15,000 of the world's biggest YouTube channels, and tracked exactly how much money this strange new content category actually pulls in.
The number they landed on: $117 million a year.
That's not a hypothetical. That's not a scare figure cooked up for a headline. That's real, trackable ad revenue flowing into a corner of the internet most people don't even have a name for yet.
1. The $117 Million Number Nobody Expected
The figure comes from a wide-reaching investigation by the video-editing company Kapwing, which set out to answer a simple question: how much of what YouTube pushes to new viewers is actually AI-generated junk, and how much money is it quietly generating?
The answer surprised even the people who ran the study. Channels built entirely around low-effort, AI-generated video — collectively nicknamed "AI slop" — are pulling in an estimated $117 million annually, and that figure only accounts for the channels researchers were able to identify and measure. The real number, almost certainly, is higher.
What makes this figure land differently than your average tech headline is the scale of normalcy behind it. This isn't a handful of viral outliers cashing in on a fluke. It's a structured, repeatable, increasingly professionalised production pipeline — one with its own communities, its own best practices, and its own economics.
2. Inside the Kapwing Study: 15,000 Channels, One Ugly Pattern
To get a fair read on the problem, researchers didn't just study whatever showed up in their own personal feeds. They pulled data from 15,000 of the most-watched YouTube channels worldwide, including the top 100 channels in each individual country. That's a genuinely global sample — not a Silicon Valley bubble, not a single-language dataset, but a true cross-section of what the world is actually watching.
Here's what stood out immediately:
- ✅ Over 20% of videos recommended to brand-new YouTube users fell into the AI slop category
- ✅ Roughly one-third of a new user's feed qualified as "brainrot" — a broader bucket of low-effort, high-repetition, attention-farming content
- ✅ 278 of the surveyed channels were producing AI slop exclusively, with no other content mixed in
- ❌ None of this required the viewer to search for or follow anything related to AI content
That last point matters most. Nobody is opting into this. YouTube's recommendation engine is doing the opting-in for them.
3. A Brand New YouTube Account Tells the Real Story
Numbers pulled from existing channels are one thing. But the researchers wanted to see what the algorithm actually does to someone starting from zero — no watch history, no subscriptions, no signals of any kind.
So they made a fresh account and watched what happened.
Of the first 500 videos recommended to that blank-slate account, 104 were AI slop. That's roughly one in every five videos shown to a brand-new user, before that user has expressed a single preference. It wasn't buried in page two of search results or tucked into some obscure corner of the platform — it was front and centre, part of the default onboarding experience for anyone opening YouTube for the first time in 2026.
Small paragraphs aside, this is the part that deserves to sit with you for a second: the platform's own machinery is choosing to introduce new users to this content, unprompted, at scale, before it has any idea what they actually want to watch.
4. 278 Channels That Run on Nothing But AI Slop
Inside that 15,000-channel dataset, researchers isolated a smaller, purer group: 278 channels that were built exclusively around AI-generated video, with zero traditional content mixed in.
Individually, these might look like small players. Collectively, the numbers are staggering:
- ✅ More than 63 billion combined views
- ✅ Roughly 221 million combined subscribers
- ✅ An estimated $117 million in yearly ad revenue between them
To put that in perspective, 221 million subscribers is more than the population of Brazil. This isn't a fringe experiment anymore — it's a parallel media ecosystem, running quietly alongside the creator economy most of us actually pay attention to.
5. Bandar Apna Dost: The $4.25 Million Monkey
If there's a mascot for this entire phenomenon, it's an Indian channel called Bandar Apna Dost. The premise, if you can call it that, features a hyper-realistic CGI monkey and a Hulk-style character traveling around in a helicopter shaped like a tomato.
It sounds like a joke. It is not.
The channel has racked up 2.4 billion views — more than most Hollywood franchises manage across their entire theatrical run — and researchers estimate it pulls in as much as $4.25 million a year. No writers' room. No production company. No traditional creator economy infrastructure at all. Just a prompt, a render, and an algorithm willing to push it to millions of screens.
6. Cuentos Fascinantes: The Channel With Almost 6 Million Subscribers
Based in the U.S. but publishing in Spanish, Cuentos Fascinantes holds the title for the most subscribers of any channel in the entire study: 5.95 million.
Its formula is simple and, frankly, a little unsettling once you know what's behind it: low-budget AI-generated cartoons aimed squarely at children. Bright colours, repetitive plots, soothing voice overs — all the hallmarks of content designed less to entertain and more to hold a young viewer's attention for as long as mathematically possible.
7. Pouty Frenchie: Selling Candy Forests to Toddlers
Based out of Singapore, Pouty Frenchie centres on an AI-generated French bulldog wandering through candy-coloured fantasy worlds — forests made of sweets, oceans made of syrup, all set against the sound of children laughing in the background.
It's earned nearly $4 million a year off roughly 2 billion views, according to the Kapwing data, and it's aimed, unmistakably, at very young viewers who have no ability to distinguish AI-generated content from anything else they're watching.
8. The AI World: Turning Real Floods Into Content Farming
Not every channel in the dataset is harmless nonsense. Pakistan's The AI World produces AI-generated short films depicting flood disasters, with titles like "Poor Family" and "Flood Kitchen," typically layered over calm, sleep-adjacent background music.
It's one of the more cynical entries researchers flagged — real human suffering, reduced to an aesthetic, generated on demand, and monetized through the exact same ad infrastructure used for cooking tutorials and product reviews.
9. Spain and South Korea: The Unexpected Capitals of Slop
This isn't a story confined to any one country or language. The data shows genuinely global reach, with two countries standing out for very different reasons:
- ✅ Spain leads the world in subscribers, with 20.22 million people — nearly half the country's population — following trending AI slop channels
- ✅ South Korea leads in raw engagement, with trending slop channels racking up 8.45 billion combined views
Content this disconnected from any specific culture or language travels easily. A video of a pressure cooker exploding, or a bulldog wandering through a candy forest, doesn't need subtitles. It doesn't need context. It just needs an algorithm willing to push it — and right now, plenty are.
10. Who Are the "Sloppers," and Why Do They Do It?
The people behind these channels have a name for themselves: "sloppers." According to reporting cited in the Kapwing findings, many of them coordinate through Telegram and Discord groups, trading prompt formulas, thumbnail tricks, and tips on gaming recommendation algorithms the way earlier generations of creators once traded SEO advice on forums.
For creators in countries like India, Nigeria, Brazil, and Ukraine, this isn't a side hustle or a joke — it's a legitimate income stream, and often a better one than what's locally available. When AI-generated ad revenue outpaces the median local wage, the incentive to keep producing isn't just present. It's overwhelming.
11. Why YouTube Has Zero Financial Incentive to Stop This?
Here's the uncomfortable truth sitting underneath every AI slop headline: this content isn't a bug in YouTube's system. It's compatible with it — arguably even optimised for it.
Long-form creators on YouTube take home 55% of net ad revenue on their watch pages. Shorts creators — the format where most AI slop lives — take a smaller cut, split from a shared Creator Pool based on view share. That structural difference means YouTube keeps a larger share of ad revenue on exactly the format where AI slop concentrates.
Run the numbers and the incentive becomes obvious. Analysts modelling the Kapwing findings estimate that $117 million flowing to creators across 63 billion views works out to roughly $1.86 per thousand views on the creator side — which implies something closer to $260 million in gross advertising revenue generated against those same views, with YouTube retaining an estimated $143 million of that itself, according to analysis published by Business Model Analyst.
Meanwhile, the platform's overall ad business isn't struggling under the weight of low-quality content — it's thriving:
- ✅ YouTube's global ad revenue climbed roughly 27%, reaching an estimated $59.36 billion
- ✅ Volume of content uploaded continues to rise year over year, largely fueled by generative tools
- ✅ Engagement metrics — the ones platforms actually optimize for — don't distinguish between a heartfelt documentary and a monkey riding a tomato helicopter
- ❌ There's no meaningful financial penalty tied to a video being AI-generated, low-effort, or repetitive
When a spokesperson says the platform is focused on "connecting users with high-quality content" while quietly removing only what violates specific guidelines, that's not spin exactly — it's a fairly accurate description of a system built to police rule violations, not quality. AI slop, by design, tends to stay just inside the lines.
12. "Workslop": The Same Problem Just Walked Into Corporate America
Here's where the story stopped feeling like a YouTube problem and started feeling like an everywhere problem.
A separate study from BetterUp and the Stanford Social Media Lab identified a corporate cousin of AI slop, one researchers have started calling "workslop" — AI-generated work output that looks polished and complete but is functionally hollow underneath. Think reports padded with generic language, emails that sound professional but say nothing, slide decks assembled by a model that never actually understood the assignment.
The study found that roughly 40% of workers had received workslop from a colleague within the past month alone. Jeff Hancock, founding director of the Stanford Social Media Lab, told CNBC there are recognizable tells once you know to look for them — purple, overwrought prose, oddly specific-sounding but ultimately meaningless word choices, and information that reads as complete while actually leaving out anything load-bearing, as reported by Yahoo Finance.
The financial impact isn't hypothetical, either. Researchers pegged the average cost of workslop-related cleanup — the time spent by human employees re-doing, verifying, or untangling AI-generated deliverables — at roughly $9 million annually for a mid-sized organization. Multiply that across every company quietly encouraging employees to "use AI to move faster," and the aggregate cost starts looking less like a rounding error and more like a hidden tax on productivity itself.
What ties this back to the $117 million YouTube figure isn't the dollar amount — it's the underlying pattern:
- ✅ Volume gets rewarded before quality gets checked
- ✅ The people producing the content face little short-term downside
- ✅ The people consuming or reviewing it absorb the real cost, just later and less visibly
- ❌ Neither system was built with a meaningful quality filter baked in from the start
Slop, in other words, isn't a YouTube problem. It's what happens anywhere generative tools meet an incentive structure that rewards output over substance.
13. The Great AI Purge That Costs Platforms Nothing
If you've followed tech headlines over the past year, you've probably seen some version of this story: a platform "cracking down" on AI slop. Spotify pulling tens of millions of low-quality tracks. Google terminating tens of thousands of account clusters. LinkedIn rolling out new detection tools. On the surface, it looks like accountability finally catching up with the problem.
Look closer at the actual mechanics, and a different picture emerges.
Spotify, for example, removed more than 75 million spammy or AI-generated tracks over a twelve-month stretch. That sounds dramatic until you look at how music royalty pools actually work: Spotify pays out a fixed pool — around $11 billion — split proportionally among the tracks that remain. Deleting 75 million low-quality tracks doesn't add a single dollar to that pool. It just changes the denominator, meaning the human artists left in the pool take a marginally larger slice of the exact same total amount. The platform's own cost from the original flood — storage and ingestion on roughly 100,000 files a day — barely registers against a business pulling in roughly 30% of global recorded music revenue.
The same logic extends across the industry. Google reportedly described terminating around 50,000 account clusters covering 130,000 channels over a six-month window — a real, measurable action, but one that, structurally, costs the platform almost nothing while generating a wave of positive "we're taking this seriously" headlines.
That's the uncomfortable throughline connecting every part of this story:
- ✅ The flood of AI slop cost platforms next to nothing to host
- ✅ The cleanup, when it happens, costs them next to nothing to execute
- ✅ Ad revenue keeps climbing on both sides of the purge
- ❌ None of it meaningfully changes the underlying incentive that created the flood in the first place
Whatever AI slop is doing to the internet — flooding feeds, draining attention, muddying trust in what's real — it hasn't done much of anything to the money. If anything, the money keeps climbing right alongside it.
14. The Illusory Truth Effect: How Fake Content Rewires Real Perception?
There's a well-documented psychological phenomenon called the illusory truth effect: the more often you're exposed to a claim, the more true it starts to feel, regardless of whether it actually is. Researchers studying the AI slop wave say this effect doesn't require the content to be convincing. It just requires repetition.
Media researcher Eryk Salvaggio has argued that AI slop functions less like entertainment and more like noise — content that drowns out legitimate signals and quietly increases how dependent people become on algorithms to make sense of what's real, a dynamic reported in coverage of the Kapwing findings. The concern isn't that any single video of a bulldog in a candy forest fools anyone. It's cumulative. Scroll past enough synthetic content, enough uncanny faces, enough "poor family" disaster shorts set to lo-fi music, and the line between real and generated starts to blur — not because any one piece was convincing, but because the sheer volume wears down your instinct to check.
A few patterns researchers point to:
- ✅ Repetition builds familiarity, and familiarity gets misread as credibility
- ✅ Emotionally simple content (cute animals, disaster footage, children's cartoons) bypasses critical evaluation faster than complex claims do
- ✅ Younger viewers, with less developed media literacy, are disproportionately exposed through kid-targeted slop channels
- ❌ Platforms currently have no meaningful labeling requirement forcing this content to identify itself as AI-generated at the point of viewing
That last point is the one regulators are starting to circle. The EU's AI Act includes disclosure provisions aimed at exactly this gap, and pressure is mounting on U.S. platforms to adopt something comparable — though as of now, most AI slop reaches viewers with zero indication of how it was made.
15. Human Creators Can't Compete With a Machine That Never Sleeps
Talk to any working YouTube creator right now and you'll hear some version of the same frustration: the platform's algorithm doesn't know the difference between a video that took three weeks to research, film, and edit, and one that took a prompt and twelve minutes of rendering time. It just knows which one people watched longer.
That math breaks in one direction, structurally. A slop channel can publish multiple videos a day, test dozens of thumbnail and title variations, and iterate on whatever the algorithm rewards in near real-time. A human creator making a single video a week simply cannot out-produce that volume, no matter how good the underlying content is.
This dynamic isn't confined to YouTube, either. Reporting on the broader creator economy has found that AI-generated image and video content has become genuinely lucrative on platforms like TikTok — enough so that it's actively pulling creators in lower-income countries toward AI slop production, specifically because it can out-earn local median wages by a wide margin, a pattern documented across creator-economy research into AI slop's spread. One frequently cited case involved a medical student in India who reported earning several thousand dollars a month from low-effort AI-generated images, a figure that dwarfs many entry-level professional salaries in that market.
None of this makes the individual "slopper" the villain of the story. Most are responding rationally to an incentive structure that rewards exactly this behavior. But it does mean human creators — the ones actually researching, filming, and editing original work — are competing against a production model with a fundamentally different cost structure. That's not a fair fight, and right now, no platform has announced a plan to make it one.
16. The Billion-Dollar Deal That Just Made Slop Mainstream
If AI slop still sounds like a fringe phenomenon confined to obscure channels, one recent deal should put that idea to rest for good.
Disney recently entered a three-year licensing partnership with OpenAI, backed by a reported $1 billion investment, that allows OpenAI's Sora video tool to generate short, prompt-based clips using hundreds of officially licensed characters, costumes, props, and environments from Disney, Pixar, Marvel, and Star Wars, according to reporting on the deal's terms described in coverage of the AI content economy. That's not a scrappy slop channel cutting corners — that's one of the largest media conglomerates on earth formally building generative, prompt-based content creation into its business model.
What that signals matters more than the deal itself. When a company with Disney's IP protections and brand standards decides prompt-generated content is worth backing at that scale, it stops being reasonable to treat "AI slop" as a temporary, low-quality anomaly the internet will simply outgrow. It's becoming infrastructure.
- ✅ Major platforms are financially entangled with the tools producing this content, not just hosting it passively
- ✅ Licensing deals like this normalize prompt-based generation as a legitimate production method, not a shortcut
- ✅ The line between "professional content" and "AI slop" is set to blur further as budgets and IP get attached to the same tools sloppers already use
- ❌ There's still no unified industry standard distinguishing high-effort generative work from mass-produced filler
17. Regulators Are Finally Trying to Catch Up
Governments have started paying attention, though enforcement is still catching up to the scale of the problem.
The clearest example is the EU AI Act, which includes disclosure provisions under Article 50 requiring AI-generated content to be identifiable as such in certain contexts. It's one of the first serious legislative attempts to force transparency at the point where a viewer actually encounters the content, rather than relying on platforms to self-report after the fact.
The music industry has moved in a similar direction with the DDEX disclosure standard, an industry framework designed to flag AI-involved tracks at the point of distribution, giving platforms and rights holders a consistent way to label generative content rather than treating every case as a one-off judgment call.
- ✅ The EU AI Act pushes disclosure obligations onto AI-generated content directly
- ✅ Industry-built standards like DDEX are emerging in parallel to formal legislation
- ✅ Pressure is mounting on U.S. platforms to adopt comparable labeling, even without a federal mandate
- ❌ No current U.S. law requires YouTube, TikTok, or similar platforms to label AI-generated video at the point a viewer sees it
Until that gap closes, disclosure remains mostly voluntary — which, given everything covered in Parts 1 through 3, means it remains mostly absent.
18. What Platforms Say vs. What They're Actually Doing?
Every major platform touched by this story has issued some version of the same statement: they take the issue seriously, they're building better detection tools, they're removing content that violates community guidelines. All of that is technically true. None of it addresses the incentive structure underneath.
Compare the public statements to the actual mechanics laid out earlier:
- ✅ Public statement: "We focus on connecting users with high-quality content"
- ❌ Actual mechanic: Recommendation algorithms optimize for watch time and engagement, not quality, and AI slop performs well on both
- ✅ Public statement: "We've removed millions of low-quality tracks/videos/accounts"
- ❌ Actual mechanic: Removals cost platforms almost nothing and don't touch the underlying payout structure that made slop profitable in the first place
- ✅ Public statement: "AI is just a tool, like any other"
- ❌ Actual mechanic: That tool now produces a meaningfully different cost structure than human-made content, and platforms haven't adjusted payouts or discovery to account for it
None of this requires assuming bad faith from any single platform. It's simpler than that — these companies are optimizing for the metrics they've always optimized for, and AI slop happens to be extremely good at hitting those metrics. Fixing that would mean changing what gets rewarded, not just what gets removed.
19. How to Spot AI Slop Before It Spots You?
You don't need a research team to start noticing this stuff. A few consistent tells show up across almost every slop channel in the Kapwing dataset:
- ✅ Titles that sound emotionally loaded but oddly generic — "Poor Family," "Amazing Story," "You Won't Believe"
- ✅ Faces or animals that move almost naturally, with small, uncanny inconsistencies in lighting or motion
- ✅ Background music that's calm or upbeat regardless of subject matter, including disaster or hardship content
- ✅ Extremely high upload frequency — multiple videos a day from a single channel, often on unrelated topics
- ✅ Comment sections that feel oddly generic, repetitive, or disabled entirely
- ❌ A polished thumbnail is not a reliable signal either way — plenty of slop channels have professional-looking thumbnails, and plenty of small human creators don't
The single most useful habit, honestly, is just pausing before you let the next video autoplay. Slop is built to work on momentum. Interrupt the momentum, and most of it loses its grip almost immediately.
20. What This Economy Means Going Forward?
Here's where I've landed after going through all of this: AI slop isn't a temporary glitch waiting to be patched out. It's a fully formed economic system, with its own production communities, its own income incentives, and — as the Disney-OpenAI deal shows — increasingly serious financial backing from companies that used to represent the gold standard of quality content.
The $117 million figure that started this whole investigation isn't the ceiling. Given rising ad revenue, expanding generative tools, and licensing deals pulling major IP holders into the same production pipeline sloppers already use, it's far more likely to be the floor.
For viewers, the responsibility currently sits almost entirely on individual attention and skepticism — there's no meaningful labeling system doing that work for you yet. For creators, the playing field remains genuinely uneven, and no platform has announced a structural fix. And for anyone hoping this gets sorted out by regulation alone, the honest answer is: not yet, and not soon enough to matter for what's showing up in your feed tonight.
Frequently Asked Questions (FAQ)
What exactly is "AI slop"?
AI slop refers to low-effort, mass-produced video or image content generated primarily with AI tools, designed to maximize views and ad revenue rather than convey meaningful information or creativity.
How much money does the AI slop economy actually make?
Research from Kapwing estimates roughly $117 million in annual ad revenue across 278 channels identified as running entirely on AI-generated content, based on a survey of 15,000 of the world's most popular YouTube channels.
Is AI slop against YouTube's rules?
Not inherently. YouTube's guidelines target specific violations like spam, misleading content, or harmful material — being AI-generated isn't, on its own, a rule violation, which is part of why this content continues to spread largely unchecked.
Which countries watch the most AI slop content?
Spain leads in subscribers, with over 20 million people following trending AI channels, while South Korea leads in total views, with trending slop channels amassing over 8 billion combined views.
Can I tell if a video is AI-generated?
Often, yes. Watch for slightly unnatural motion, generic emotionally-loaded titles, mismatched music and subject matter, and unusually high daily upload frequency — all common patterns across the channels researchers identified.