YouTube still pushing eating disorder content to teens despite new UK safety rules

A year after the UK introduced strict new rules meant to protect teenagers from dangerous online content, YouTube's recommendation system is still serving up videos that glorify eating disorders to young users. Researchers at the Centre for Countering Digital Hate set up a test account posing as a 13-year-old girl in the UK, watched ten videos about dieting and body image, and then tracked what the platform's algorithm suggested next. Out of the next hundred recommendations, one in ten featured material classified as harmful—content promoting extreme thinness, severe calorie restriction, or other dangerous practices. The improvement is real: two years earlier, the same experiment found one in four recommendations were harmful. But for researchers and advocates, incremental progress feels hollow when any amount of this content still reaches vulnerable teenagers.

The legal framework driving this scrutiny arrived in July 2025, when the Online Safety Act took effect. The law gives platforms like YouTube a direct legal obligation to protect people under eighteen from content that encourages eating disorders, self-harm, or suicide. It also requires companies to examine how their algorithms might cause harm and to take steps to reduce that risk. The stakes are substantial: Google, which owns YouTube, could face fines up to ten percent of its global revenue—potentially billions of pounds—if it fails to comply. Yet the CCDH's research suggests the company's efforts remain insufficient.

The videos still slipping through are explicit in their danger. One account compiles images of girls idealizing extreme thinness. Another promotes a diet of just 170 calories per day, far below what a teenager needs. A third claims to help users lose weight through subliminal messaging and links to a document promising "the most emaciated skeletal dainty body eva." None of these videos triggered YouTube's crisis panel—a blue box that appears beneath sensitive content and directs users to support services. Other diet and body image videos that the researchers did not classify as harmful did generate these warnings, suggesting the system is inconsistent in identifying what actually poses risk.

Google has responded by saying it prohibits content that encourages eating disorders while allowing recovery stories, and it has removed the specific videos cited in the CCDH report for violating community guidelines. The company points to partnerships with the NHS, Mind, and other organizations, and to curated expert videos that appear when teenagers search for topics like depression or eating disorders. But these measures have not prevented the algorithm from continuing to recommend dangerous material to users who show early interest in diet and body image content.

The human cost of this gap between policy and practice is visible in the story of Jazmin Kaur, now 22, who was diagnosed with anorexia at thirteen. She spent six years in NHS treatment, a period during which she was constantly online, consuming material that made her condition worse. "Every time I left hospital, I'd have my phone with me and I'd constantly be on it," she said. The content she encountered was so extreme that she internalized it as reflecting her own vulnerabilities. She eventually deleted her social media accounts entirely and is now pursuing a master's degree in paediatric nursing while working in mental health services. Her experience underscores what researchers mean when they say that for vulnerable users, a small algorithmic nudge can push someone into a dangerous situation.

Victoria Longley, chief executive of the eating disorder charity Beat, acknowledges that online communities can sometimes provide valuable support and reduce isolation, particularly when NHS access is limited. But she notes that roughly ninety percent of people her organization speaks to report encountering harmful content online. The charity offers practical advice—turning off notifications, using time-limiting apps, blocking accounts—but Longley is clear that responsibility ultimately rests with the platforms themselves to ensure their systems are safe.

Alexandra Johnson, a senior research manager at CCDH, framed the findings with cautious optimism. The fact that harmful recommendations dropped from one in four to one in ten shows that regulation does have an impact. "But one video is too many," she said, "and we don't want any of this content to get through, particularly to vulnerable users." The research was replicated in the US and EU with similar results, suggesting the problem is not unique to the UK.

The government is moving toward more aggressive intervention. In June, it announced plans to prevent anyone under sixteen from accessing major platforms including YouTube, TikTok, Instagram, Snapchat, Facebook, and X, with these restrictions expected to take effect in spring 2027. Whether age gates will prove effective remains uncertain, but the direction is clear: regulators have lost patience with the idea that platforms can police themselves through algorithms and content moderation alone.