A new investigation has raised fresh questions about YouTube’s recommendation systems after finding that young children are being shown large volumes of AI-generated videos presented as educational material. The issue appears to sit at the intersection of platform design, disclosure rules and the growing economics of low-cost synthetic content.
A new investigation by The New York Times, reported by The Verge, found that YouTube is recommending substantial amounts of AI-generated video to children who have been watching established children’s content such as CoComelon, Bluey and Ms Rachel. According to that reporting, more than 40 per cent of the Shorts recommended after those viewing sessions appeared to contain AI-generated visuals, suggesting that synthetic material is not peripheral to children’s feeds but is being surfaced in a systematic way by the platform’s recommendation engine.
The material in question is not being presented as experimental AI media. Rather, it is often packaged as suitable for toddlers and pre-school children, with videos claiming to teach animals, colours, songs or other basic concepts. Earlier and parallel reporting from WIRED has shown that child-directed AI videos on YouTube frequently rely on bright animation, familiar nursery-rhyme conventions and child-friendly descriptions, even when the underlying content is poorly made, incoherent or plainly misleading. The concern is less about the existence of AI-made clips than about their presentation as trustworthy educational content.
That matters because YouTube continues to market YouTube Kids as a more controlled environment for children. In its own materials, the company says the app is built for children aged 12 and under and offers parents a choice of content settings, including the option to approve content manually. YouTube also says parents can block search, limit screen time, remove individual videos or channels, and rely on age-based viewing categories. On paper, that presents YouTube Kids as a filtered alternative to the main platform rather than merely a simplified front end.
Yet the present dispute highlights a structural weakness in that model: parental controls do not necessarily solve the problem if the wider recommendation and classification systems are already allowing low-quality synthetic content to be treated as suitable for children. WIRED noted in earlier reporting that many parents still use the main YouTube app rather than the separate children’s app, and that the platform has long faced criticism for hosting videos that look child-friendly at first glance but prove unsuitable on closer inspection. The current AI wave appears to reproduce that problem in a newer and cheaper form.
A central policy gap concerns labelling. YouTube’s official rules require creators to disclose altered or synthetic content when it appears realistic, and the platform says it may in some cases add a label itself. However, the company’s published guidance also makes clear that unrealistic or animated content does not generally require disclosure. That distinction is important because much of the AI-made children’s material now drawing criticism is obviously artificial, yet still capable of being mistaken for harmless educational entertainment. In effect, some of the content most likely to reach children may fall outside the disclosure system designed for realistic deepfakes.
YouTube has nevertheless been taking some enforcement action. Reporting on the latest New York Times findings said that, after examples were shared with the company, five cited channels were suspended from the YouTube Partner Programme. That sits alongside YouTube’s broader monetisation rules, which state that repetitive or mass-produced “inauthentic” content is ineligible for monetisation. The platform tightened the wording of that policy in July 2025, saying the change was intended to clarify long-standing restrictions on spam-like or low-effort uploads rather than introduce an entirely new rule.
The company has also acknowledged the wider “AI slop” problem. In his 2026 letter on the future of the platform, YouTube chief executive Neal Mohan said the rise of AI had increased concern about low-quality content and that the company was building on its existing anti-spam and anti-clickbait systems to reduce the spread of repetitive, low-quality AI material. That statement indicates that YouTube recognises the scale of the issue, but it also underlines how reactive the platform still appears to be: the recommendation machinery is already distributing content that the company now says it wants to suppress.
The issue is also relevant in Europe, where child safety on large platforms is already under regulatory scrutiny. European Commission is examining safeguards for minors on YouTube and other major services under the Digital Services Act, including questions around age verification and the prevention of harmful material reaching children. That followed an earlier 2023 EU information request to YouTube and TikTok on the protection of minors. The latest reporting on AI-generated children’s videos is therefore likely to feed into an existing regulatory debate.
For YouTube, the difficulty is not simply whether individual videos breach policy, but whether the platform’s incentives, labelling rules and recommendation systems are together creating a route by which synthetic, low-quality material can be pushed to very young audiences under the cover of educational content. Parents can use stronger controls, including approved-content-only settings, but the burden still appears to fall heavily on families to distinguish reliable children’s programming from content produced quickly and cheaply at scale. The latest findings suggest that this is no longer a marginal moderation issue, but a test of whether YouTube’s child-safety framework is keeping pace with generative AI.
First published on eutoday.net.
