Teenagers today don’t need to go looking for violent videos online. Increasingly, the videos find them first. One minute a young person is watching football clips or dance trends, the next they are being served footage of street fights, stabbings or “postcode war” videos by an algorithm that has quietly decided: “You watched one. Fancy ten more?”
Social media addiction
Social media platforms are built to keep users scrolling. The longer someone stays on the app, the more adverts they see, the more data is collected, and the more profitable the platform becomes. Algorithms are designed to learn what grabs attention and unfortunately, outrage, shock and violence are very good at just that.
The dark side of the “For You” page
Research into recommendation systems on platforms like TikTok, Instagram Reels and YouTube Shorts has found that harmful and distressing content can quickly become amplified to young users through automated recommendations. A recent study examining 3,000 videos shown to child accounts found that harmful content appeared within just three minutes of passive scrolling on YouTube for younger users.
Violent algorithms
The scale of the problem is alarming. Research from the Youth Endowment Fund found that 70% of teenagers in England and Wales had seen real life violent content online in the past year. Crucially, only 6% said they actively searched for it, whilst 25% said it was promoted to them through algorithmic features such as “For You” pages, feeds and recommendations.
TikTok shock
TikTok was identified as the platform where teenagers were most likely to encounter violent material, with 44% of Teenage TikTok users reporting exposure to violent content.
Normalisation of extreme behaviour
For teenagers, whose brains are still developing and who are naturally drawn to risk, status and belonging, this creates a dangerous cycle. Violent videos are not just consumed, they are normalised. Repeated exposure can make extreme behaviour seem ordinary and is feeding into the desensitisation of a generation of children. And the algorithms? They are not neutral. They are a digital feedback loop with absolutely no moral compass.
Machine learning
Evidence increasingly suggests these systems learn frighteningly quickly. Researchers studying TikTok’s recommendation engine found that content reinforcement can occur within the first 200 videos watched, with algorithms rapidly narrowing what users are shown based on tiny engagement signals such as pauses, rewatches and likes.
Online Safety Act
This is where the UK’s Online Safety Act comes in. It was introduced to tackle exactly this. Under the legislation, platforms likely to be accessed by children must assess risks and take steps to protect young users from harmful content, including violent material, bullying and dangerous behaviours. Ofcom now requires platforms to introduce stronger safety measures, age checks and systems to reduce children’s exposure to harmful content.
If you profit, you are responsible
That is a significant step forward. For years, social media platforms have cultivated plausible deniability by treating harmful content as “user behaviour” rather than as an outcome shaped by the platform architecture itself. The Online Safety Act finally establishes a new principle: if you profit from the platform, you are responsible for what happens on it.
The law is only as strong as its enforcement
Critics argue that many age-verification systems are easy to bypass, with children reportedly using fake birthdays, borrowed accounts and even makeup tricks to fool age checks. Recent research found that nearly a third of UK children had successfully bypassed age-verification systems, while almost half believed such checks were easy to get around.
Viral violence
Meanwhile, harmful content still spreads rapidly before moderators can remove it. Platforms often react after damage is already done, rather than preventing it from reaching children in the first place. Ofcom research has also found that while 67% of young people encounter harmful content online, fewer than one in five actually report it, often because they believe nothing will happen anyway.
Legislation needs to go further
If a platform’s algorithm repeatedly pushes violent content to children, there should be serious consequences, not just polite warnings and strongly worded letters. Tech companies should face meaningful fines, mandatory transparency around recommendation systems, and independent audits showing how harmful content spreads.
Parental controls
Parents cannot compete with billion-pound algorithms engineered by teams of behavioural scientists. Asking families alone to “monitor screen time” is a bit like handing someone a bucket while Silicon Valley floods the house.
Toxic trends in the Tees Valley
Social media companies also need to stop hiding behind the phrase “user-generated content”. Algorithms actively promote content. They decide what trends, what spreads and what goes viral. If an app can recommend trainers you looked at once for the next six weeks, it can certainly stop recommending knife-fight videos to a 14-year-old.
This issue feels particularly urgent in places like the Tees Valley, where communities already face the devastating impact of knife crime and violent offending. Knife crime increased from 2024 to 2025, with 155 offences per 100,000 people seeing the area as the worst affected by knife crime in the entire country.
Social media is undermining anti-violence efforts
For some teenagers, online violence becomes part of everyday culture that is shared in group chats, reposted for clout and discussed like football highlights. In areas like the Tees Valley, where communities are already working hard to steer young people away from crime, algorithms pushing violent content risk undermining that work completely.
Digital Wild West
The internet does not have to be a digital wild west where children are left to dodge violent content like online tumbleweeds. The technology exists to create safer spaces. The question is whether governments are willing to force platforms to use it and whether platforms are willing to put children’s safety ahead of engagement statistics.
Because “the algorithm made me do it” should never become society’s excuse for failing a whole generation of children.
