Your toddler is sitting in front of the tablet. The video looks harmless — bright colors, a bouncy jingle, cartoon characters singing about cars or letters. You glance over, see nothing alarming, and go back to making dinner.
Here's what you didn't catch: the cartoon children in that "nursery rhyme" video are riding without seat-belts, climbing out of a moving car, walking through traffic. No human wrote that scene. No child-safety expert reviewed it. A machine generated it in minutes, and it's already been watched tens of thousands of times.
We dug through the investigations, the expert letters, and the platform data to find out exactly how deep this problem goes. Here are 17 truths every parent needs to see.
📋 Table of Contents
- AI Videos Are Teaching Toddlers to Ignore Seat-belts and Walk Into Traffic
- Choking and Poisoning Hazards Are Being Modelled as "Educational" Content
- Genuinely Frightening Imagery Is Slipping Into Toddler Content
- One Channel Can Out-Produce an Entire Studio in Days
- "Educational" AI Videos Are Often Just Factually Wrong
- Experts Say This Isn't "Brain Rot" — It's "Brain Stunt"
- There's Real Money Fuelling the Flood
- 200+ Experts Had to Send a Formal Warning Letter to Google's CEO
- AI Content Has Already Breached YouTube Kids
- The Algorithm Itself Is Spreading the Problem
- Even Brand-New Accounts Get Served AI Slop
- Fake "Educational Science" Content Isn't New — It's Just Multiplying
- Elsagate Never Really Ended — AI Just Supercharged It
- AI Is Generating Disturbing Content in Kid-Adjacent Spaces Too
- Some Channels Run 10+ Hour AI Live-streams to Entrance Infants
- YouTube's Enforcement Is Reactive, Not Preventive
- The Fixes Experts Are Demanding Still Haven't Happened
1. AI Videos Are Teaching Toddlers to Ignore Seatbelts and Walk Into Traffic
Investigative reporting from The 74, co-published with Mother Jones, uncovered a "nursery rhyme" video about cars where two cartoon children ride without seat-belts and walk in the middle of a road with moving traffic behind them.
The video has been played almost 50,000 times. It's styled exactly like the kind of content a parent would trust — bright, musical, kid-friendly — which is exactly why it's dangerous.
- ✅ Marketed as safe, sing-along content
- ✅ Modelled real unsafe behaviour a toddler could imitate
- ❌ No human safety review before publishing
2. Choking and Poisoning Hazards Are Being Modelled as "Educational" Content
It gets worse than bad driving habits. Reporters documented AI-generated videos showing a crawling baby biting into a whole grape — a textbook choking hazard for that age group — along with a baby eating honey, which carries a real risk of infant botulism, and a video of a teacher eating raw, uncooked elderberries, which are toxic.
None of this was flagged as satire or fiction. It was presented as gentle, everyday toddler content.
- ✅ Whole grapes shown as safe snacking
- ✅ Honey fed to a crawling infant on screen
- ✅ Raw elderberries eaten with no warning
3. Genuinely Frightening Imagery Is Slipping Into Toddler Content
Beyond physical danger, some AI videos cross straight into psychological distress. Children's media consultant Carla Engelbrecht — who has built digital experiences for Sesame Street, PBS Kids, and Highlights for Children — has spent months documenting this content and says the volume is staggering.
Among her findings: a video of a scared child being chased by a T-Rex, rendered with unsettling AI realism rather than the softened, age-appropriate style a real children's studio would use.
"The more content I find, the more horrified I get," Engelbrecht told Undark.
4. One Channel Can Out-Produce an Entire Studio in Days
The channel Jo Jo Funland, creator of the viral "Vroom Vroom! Car Ride Song," posted more than 10,000 videos in just seven months — an average of roughly 50 new uploads every single day.
For scale: Sesame Street has published fewer than 4,000 videos to YouTube across its entire 20-year history on the platform. One AI-driven account matched two decades of a trusted studio's output in under a year.
- ✅ 10,000+ videos in 7 months
- ✅ ~50 new uploads daily
- ❌ Zero editorial review at that pace
5. "Educational" AI Videos Are Often Just Factually Wrong
A huge share of this content brands itself as educational — ABCs, counting, geography. The problem is that AI-generated "educational" videos are frequently, quietly, wrong.
Reporters reviewing this wave of content found sing-along videos where letters appear garbled mid-frame, and geography videos where U.S. state names come out mispronounced or misspelled entirely. A toddler watching on repeat has no way to know the difference between a real lesson and a hallucinated one.
- ✅ Letters that morph or garble on screen
- ✅ State names spelled incorrectly in "learning" videos
- ❌ No fact-checking layer before publishing to millions of views
Keri Ewart, an associate professor of education at Wilfrid Laurier University who researches AI literacy, put it simply to CBC News: this kind of content "doesn't really have any developmental, cognitive coherence or intention to it."
6. Experts Say This Isn't "Brain Rot" — It's "Brain Stunt"
Adults call low-quality AI content "brain rot." Dana Suskind, professor of surgery and paediatrics and co-director of the TMW Center for Early Learning, argues that term doesn't go far enough for toddlers.
Her reasoning: adult brains are already formed, so junk content just dulls them temporarily. A toddler's brain is still being actively wired. Feeding it error-riddled, plot less AI content during that window isn't rot — it's active interference with development itself. She calls it "brain stunt."
Suskind and other specialists point to several specific developmental risks:
- ✅ Reduced ability to distinguish reality from fiction
- ✅ Shortened attention spans from over-stimulating pacing
- ✅ Disrupted language and social development tied to excess passive screen time
- ✅ Direct imitation of unsafe behaviours modelled on screen
"It's arriving during a critical developmental period, when children's brains are being wired, truly, for a lifetime," Suskind said.
7. There's Real Money Fueling the Flood
None of this is happening by accident. It's a business model.
A letter organised by the child-safety nonprofit Fairplay and sent directly to Google and YouTube leadership states plainly that the top-watched AI slop channels targeting kids have earned over $4.25 million in annual revenue. Some creators have openly bragged about the money flowing in from plot-less, mesmerising AI content aimed at an audience too young to tell the difference between real programming and mass-produced filler.
- ✅ $4.25 million+ in annual revenue for top AI kids channels
- ✅ Low production cost, near-infinite output
- ❌ An audience that can't distinguish quality or push back
Low cost to produce, high watch-time from an audience that can't scroll away on its own, and monetization built into the platform — it's a formula that rewards volume over quality every time.
8. 200+ Experts Had to Send a Formal Warning Letter to Google's CEO
When advocacy and awareness weren't moving the needle, more than 200 child specialists, advocacy groups, and schools escalated the issue directly to the top.
The letter — addressed to Google CEO Sundar Pichai and YouTube CEO Neal Mohan, and signed by organisations including the American Federation of Teachers and the National Black Child Development Association — didn't mince words:
- ✅ Called AI content "toddler misinformation," not benign entertainment
- ✅ Accused YouTube of prioritising profit over the wellbeing of its most vulnerable users
- ✅ Demanded clear labelling of all AI-generated content on the platform
- ✅ Demanded a ban on AI-generated videos within YouTube Kids specifically
That it took a formal, multi-organisation letter to a sitting CEO — rather than routine platform moderation catching the problem — says a lot about the scale involved.
9. AI Content Has Already Breached YouTube Kids
Here's the part that should worry parents most directly. Many families specifically choose the YouTube Kids app because it's marketed as a curated, safer space than the main platform.
Most of the disturbing videos identified in early investigations were found on regular YouTube rather than YouTube Kids. But that separation is already breaking down — reporting from the New York Times found AI-generated videos had made their way into YouTube Kids as well, despite the app's promise of stricter curation.
- ✅ YouTube Kids marketed as curated and safe
- ✅ AI slop confirmed present in main YouTube feeds
- ❌ AI content confirmed to have reached YouTube Kids too, per NYT reporting
A YouTube spokesperson has said the company has "invested deeply in age-appropriate environments" and that YouTube Kids offers "industry-leading parental controls." The on-the-ground findings suggest those controls are still catching up to the pace of AI content production.
10. The Algorithm Itself Is Spreading the Problem
This isn't just a matter of a few bad channels slipping through. YouTube's own recommendation system is actively funnelling AI slop toward kids who were watching legitimate content.
Fairplay's letter to Google states that after watching a trusted channel like Cocomelon, 40% of the videos YouTube recommended next contained AI-generated content. A parent doing everything right — starting with a known, reputable channel — can still watch their child get algorithmically redirected straight into the slop pile.
- ✅ Cocomelon used as the trusted starting point
- ✅ 40% of what came next was AI-generated
- ❌ No parental action required for the algorithm to make the switch
11. Even Brand-New Accounts Get Served AI Slop
It's not just returning viewers. The same letter found that about 21% of all new users' recommended Shorts contained AI slop.
That means a child (or a parent setting up a fresh account specifically to avoid a messy watch history) gets served this content almost immediately, with no viewing pattern to blame. The recommendation system doesn't need a history of bad choices to start pushing AI-generated content — it's simply baked into what a large share of "kids' content" on the platform now is.
12. Fake "Educational Science" Content Isn't New — It's Just Multiplying
Before AI slop had a name, the underlying problem already existed. A 2023 BBC investigation found that false science information in YouTube videos was being recommended to older kids under the "educational" label.
What's changed since then is production speed. What used to require a misinformed creator sitting down to film a video can now be mass-generated by AI in bulk, meaning the same category of problem — confidently wrong "educational" content — is being produced at a volume no team of human moderators can realistically keep pace with.
- ✅ False science content flagged as far back as 2023
- ✅ AI has since multiplied the volume, not fixed the accuracy
- ❌ No indication moderation speed has kept up with generation speed
13. Elsagate Never Really Ended — AI Just Supercharged It
Long before "AI slop" entered the vocabulary, there was Elsagate — the 2017 scandal where disturbing videos featuring beloved children's characters in violent or sexual scenarios flooded YouTube's kids' ecosystem, hidden behind innocent thumbnails and titles.
YouTube made public commitments to crack down on it. But according to Fairplay's letter, young viewers are still being targeted with disturbing or inappropriate videos featuring their favorite characters or animals in violent or sexual situations — years after the platform said the problem was handled.
- ✅ 2017 Elsagate crackdown promised to fix character-exploitation content
- ✅ Similar content patterns documented again in Fairplay's 2026 letter
- ❌ AI mass-production makes the same tactic far easier to scale than it was in 2017
What made Elsagate shocking in 2017 required real creators putting in real (if malicious) effort. AI removes that bottleneck entirely — the same manipulative tactic can now be replicated thousands of times over with a fraction of the labor.
14. AI Is Generating Disturbing Content in Kid-Adjacent Spaces Too
The AI slop problem doesn't stay neatly inside videos explicitly labeled "for kids." Fairplay's letter to Google references a 2025 WIRED investigation that found channels with names resembling wholesome, family-friendly titles like "Cute Cat" were actually filled with AI-generated videos of animals being brutally tortured.
This matters for families because channel names and thumbnails are often the only screening tool a busy parent has time to use. When a channel branded to look safe for a young audience is secretly hosting disturbing content, that filtering shortcut stops working entirely.
- ✅ Innocent-sounding channel names used as camouflage
- ✅ Thumbnails styled to look family-friendly
- ❌ Actual content far outside what the branding suggests
15. Some Channels Run 10+ Hour AI Livestreams to Entrance Infants
Fairplay's letter also flags a specific format designed for maximum passive viewing time: 10-plus-hour AI-generated live-streams built around what the letter describes as "zombifying" animations, engineered so a parent can put even an infant in front of the screen for hours at a stretch.
The letter is careful to note that no content designed to hijack a child's attention for that long is acceptable — but AI makes the tactic dramatically easier to execute, since a single machine-generated loop can run indefinitely without a human ever needing to produce new material.
- ✅ 10+ hour continuous AI live stream formats
- ✅ Marketed toward infants specifically
- ❌ Designed around attention-hijacking, not education or entertainment value
16. YouTube's Enforcement Is Reactive, Not Preventive
To be fair, YouTube hasn't ignored the problem entirely. In response to investigative reporting from The 74 and Mother Jones, the platform terminated six channels for violating its terms of service and one more for spam violations.
But look at the pattern: action only came after journalists did the investigative legwork and published their findings publicly. A YouTube spokesperson acknowledged as much, noting the platform doesn't catch every violation given the sheer volume of content uploaded.
- ✅ Six channels terminated after press investigation
- ✅ One additional channel removed for spam violations
- ❌ No indication these channels were caught before reporters flagged them
- ❌ No proactive detection system specifically built for AI kids' content described publicly
Given that one single channel, Jo Jo Funland, was uploading roughly 50 videos a day on its own, a handful of after-the-fact terminations barely scratches the surface of the total volume being produced platform-wide.
17. The Fixes Experts Are Demanding Still Haven't Happened
Fairplay's letter laid out exactly what it would take to actually fix this, not just patch it after the next news cycle:
- ✅ Clearly label all AI-generated content on YouTube
- ✅ Prohibit AI-generated content from being hosted on YouTube Kids entirely
- ✅ Prohibit AI-generated "Made for Kids" videos from the main platform as well
As of this writing, none of these three changes have been fully implemented. YouTube has pointed to existing tools like YouTube Kids' parental controls as evidence of its commitment to child safety, but the specific structural fixes experts asked for remain unaddressed.
That gap is exactly why this list exists — because until platform policy catches up, the burden of catching this content falls on parents watching closely enough to notice it in the first place.
How to Protect Your Child Right Now?
- ✅ Preview a full video yourself before letting your toddler watch unsupervised
- ✅ Watch for AI tells: garbled letters, warped faces, nonsensical plots, repetitive loops
- ✅ Turn off autoplay so the algorithm can't hand off to the next random video
- ✅ Favor channels tied to known, trusted children's media brands
- ✅ Treat "educational" claims in the title as unverified until you've watched it yourself
For a closer look at what to watch for, ABC News Live spoke with a Bloomberg tech reporter about spotting and avoiding this content: Tips on how to keep your babies away from 'AI slop'.
Frequently Asked Questions (FAQ)
Is AI slop content actually dangerous, or just low quality?
Both. Some of it is simply nonsensical and low-value. But documented cases — like videos modelling seat-belt-free car rides, whole grapes fed to crawling infants, or honey given to babies — cross into content that models genuinely unsafe behaviour a toddler could imitate.
Is this content on YouTube Kids, or just regular YouTube?
Most documented cases were found on the main YouTube platform, but New York Times reporting confirmed AI-generated videos have also reached YouTube Kids, despite its curated, "safer" positioning.
How can I tell if a kids' video was AI-generated?
Look for garbled or morphing text, inconsistent character movement, nonsensical plots, mispronounced or misspelled words in "educational" content, and unusually high upload frequency from the channel.
What is YouTube doing about this?
YouTube has terminated a small number of channels following press investigations and points to YouTube Kids' parental controls as a safety measure. Child-safety advocates say this is reactive rather than preventive, and are still waiting on platform-wide labelling and a ban on AI content within YouTube Kids.
What can parents do today?
Preview videos before your child watches unsupervised, turn off autoplay, watch for AI production tells, and favour established children's media brands over unfamiliar high-volume channels.