YouTube Still Pushes Eating Disorder Videos to Simulated Teen Accounts, One Year Into UK's Online Safety Act

Since July 2025, when the core algorithm-safety provisions of Britain's Online Safety Act took effect, YouTube has been legally required to protect minors from content promoting suicide, self-harm and eating disorders. A new test by the Centre for Countering Digital Hate shows the platform still isn't doing that job well enough.

The CCDH, a nonprofit that researches online harms, built a simulated account posing as a 13-year-old girl in the UK. Researchers fed it 10 videos about dieting and body image, the kind of content a curious teenager might click on without knowing where it leads. Then they tracked the next 100 videos YouTube's "Up Next" algorithm served up.

The result: one in 10 recommended videos featured thinspiration, extreme calorie restriction, or other harmful eating-disorder content, according to the CCDH's findings reported by the BBC. That's an improvement from two years ago, but it's still a legal violation waiting to be enforced, and it's still landing in front of a simulated child.

One Family's Real Version of the Simulation

Jazmin Kaur isn't a research data point. She's 22, from Leicester, and spent six years in NHS treatment for anorexia after being diagnosed at 13.

"It all started quite innocently," Kaur told the BBC. "I wanted to get fitter, I wanted to get healthier, so I started going online and took social media very much at face value without understanding the facts."

She said some of what she found online was genuinely helpful. Most of it wasn't. "I was fed such extreme content towards the end that I took it for my own vulnerabilities," she said. Even hospital discharge didn't break the cycle. "Every time I left hospital, I'd have my phone with me and I'd constantly be on it," she said.

Kaur eventually deleted her social media accounts at university. She's now finishing a master's degree in pediatric nursing and working weekends in an adult mental health unit. Her recovery is real, but it took years and a full disconnect from the platforms that fed her.

Eating disorders have complicated causes. No researcher, including the CCDH, claims YouTube alone creates them. But an algorithm serving up extreme content to a vulnerable 13-year-old isn't a neutral bystander either.

What the Law Actually Requires and Where It Falls Short

The Online Safety Act gives Ofcom, Britain's telecoms regulator, the power to fine companies up to 10% of global revenue for failing to protect minors from this kind of content. For Google, that threshold could mean billions of pounds.

Ofcom has already said publicly that YouTube and TikTok aren't doing enough, and it's calling for stronger protections. That's a regulator criticizing the companies it oversees, not a formal enforcement action. No fine has been levied against Google over this specific finding, and no investigation tied directly to the CCDH report has been announced.

Google's response, provided to the BBC, calls its commitment to blocking harmful content "steadfast" and says the specific videos flagged in the CCDH report have been removed. The response doesn't address why the algorithm kept recommending similar material one in 10 times in the first place.

The Fair Question About What "Enough" Means

There's a legitimate argument that content moderation at YouTube's scale, with hundreds of hours of video uploaded every minute, will never hit zero percent harmful recommendations. Perfect enforcement against every diet video, every body-image post, every borderline clip is not realistic, and companies that promise it are lying.

But a one-in-ten hit rate on a law specifically designed to stop this isn't a rounding error. It's the exact failure mode Parliament wrote the statute to prevent, sitting there a year after the deadline for compliance passed.

The design problem is real too. Algorithms are trained to maximize watch time, not minimize harm, and no outside party can audit Up Next's black-box recommendation logic in real time. That makes claims of improvement hard to verify independently, and makes YouTube's word the only word available unless regulators demand access to the system itself.

Ofcom hasn't announced a formal investigation or fine over this report. The CCDH's test used a single simulated account, not a comprehensive audit of the platform. Whether Ofcom escalates from public criticism to an actual enforcement action, using the fine authority Parliament gave it, is the open question nobody in this story has answered yet.