There's a number in EdSurge's report on AI-generated children's video that does more work than anything else in the piece, and it isn't about harm.

Do the arithmetic. That's about fifteen videos a day, every day, with no weekends. No studio works that way. No reviewer watches that much. Before you look at a single frame, the number has already told you something: nothing in that library was checked by a person.

Early childhood educators have started calling this material "baby slop."

EdSurge reports the tell that makes the point better than any statistic:

Two things are true about that bus, and both matter.

Nobody chose it. It isn't malice or a prank — it's a generative model doing what generative models do when nothing constrains them, in a clip made to be one of thousands.

And nobody caught it. That's the part that should hold your attention. In a world where someone reviewed the output, a burning school bus in a counting video is a five-second fix. In a world where 5,500 videos ship in a year, there is no one standing at that door.

Elsewhere, EdSurge reports, inaccurate information and hallucinated visuals show up routinely. As Allison Bacon, coordinator of instructional technology at the Ossining Union Free School District in New York, describes the format: "It's easy to make an eight-second AI video, and we wind up with these very quick, colorful, fun things that don't really have any meaning."

(This section is context, not EdSurge's reporting.)

Children's television used to be expensive. That single fact did more for quality control than any regulation, because when a show costs real money to produce, someone watches it before it airs — not out of virtue, but because the money makes review worth doing.

Generative video collapsed the cost of making a clip to roughly nothing. It did not change the cost of reviewing one, which is still a human being watching in real time, at real time speed.

That asymmetry is the whole story. When creation is free and review is not, review doesn't get skipped by decision — it becomes arithmetically impossible. You cannot watch fifteen new videos a day from one channel, let alone from every channel.

It's also worth remembering this isn't the first version of this problem. Around 2017, YouTube faced what came to be called "Elsagate": disturbing content dressed in the visual language of children's cartoons, surfacing in kids' feeds through recommendation. That wave was made by humans, and it still outran moderation. What's changed is that the production step no longer requires anyone at all.

Here I want to be careful, because this is where most coverage of this story goes wrong.

The educators EdSurge interviewed are genuinely concerned. Jennifer Chen, a professor of early childhood education at Kean University, on how hard this is to spot: "Some of the content may look educational and engaging. You might not realize it unless you really look at it closely."

Keri Ewart, an associate professor of education at Wilfrid Laurier University, goes further. When words don't match images, she says, "that cognitive disparity is causing children to learn inaccurately," and:

EdSurge also reports that repeated viewing may damage attention spans and lead to cognitive overload, and quotes Bacon on why the age matters: young children "are still building neurological processes. They're learning how to experience the world. Entertainment generated through AI to mindlessly babysit small people is inherently problematic."

Now the caveat, and please don't skip it. Every one of those statements is a professional judgment from an educator. None of them is a research finding. No study is cited in this report establishing that AI-generated video specifically harms child development — and realistically, the format is too new for that evidence to exist yet.

That doesn't make the concerns worthless. People who spend their days with four-year-olds notice things before researchers can measure them, and dismissing that is its own kind of foolishness. But "experienced teachers are worried" and "we know this harms children" are different claims, and only the first one is currently supported.

If you see this story told the second way — and you will — you're reading someone who either didn't check or didn't care.

What needs no study at all is the accountability problem: video aimed at preschoolers is being produced at a volume that makes human review impossible, and no one is answerable for what ends up inside it. You don't need a longitudinal study to find that unacceptable.

The most useful part of the EdSurge report is that the educators are specific. Their signals:

For vetting help, the educators point to Common Sense Media, and to the American Academy of Pediatrics' "Stop, Wait, Go" framework for families.

The best line in the article is not a warning. It's Bacon's:

That is worth more than any blocklist, because it scales in the one direction filters don't. Detection tools and curated allowlists age badly — they're in a race with the generators and they are losing it. A child who has been taught to notice keeps that skill as the fakes improve.

Bacon's broader framing is the same move: "Really push the fact that we have to be thinkers and we have to be looking with that critical lens. Ask kids, 'What did we notice? What's good? What's different?'" And when something strange does turn up: "What did we see? What did we think? The human connection and conversation should be the main focus."

For parents specifically, EdSurge describes what Josefa Martinez, a preschool language teacher at the Park School in Ossining, sends families: vetted sites she's personally watched, plus follow-up questions to ask afterwards. If a child watches something about firefighters — "How can a firefighter support our community?" If it's about numbers — "How do you know that five is bigger than two? Can you show me by using some of your toys?"

Her reasoning: "These questions make sure the child is actually engaged and not just sitting there mindlessly watching."

Which points at something the AI framing can obscure. The problem with a video nobody made and nobody watched isn't only what's in it. It's that it's designed to be consumed alone, without a person. The counter-move isn't better filtering. It's being in the room.

Ewart notes that more schools are convening "social media nights and AI evenings" so parents can learn to vet this material, and adds the thing worth ending on:

Source: "AI Videos Are Warping Early Learning, Teachers Say. Here's How They're Responding," EdSurge, reporting by Adam Stone, September 8, 2026: https://www.edsurge.com/news/ai-videos-are-warping-early-learning-teachers-say-heres-how-theyre-responding

Sections marked as context — the Elsagate history, the economics of review, and the note on text rendering — are background, not EdSurge's reporting.

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter