An unexpected recommendation can become an entire feed. Here’s how to understand what happens next, recognise when it becomes a problem, and help children regain some control.

A parent asks why their child has been watching so many videos about the same subject. The child replies, “I wasn’t looking for them. They just kept appearing.”

It’s tempting to hear that as an excuse. After all, someone must have watched the first video. Perhaps they even watched the second. But recommendation systems are designed to make decisions about what appears next, and those decisions can gradually turn a passing interest—or even an uncomfortable moment—into a much more persistent digital experience.

A child does not need to search for the same subject repeatedly to keep seeing it. The platform may do much of that work for them.

For parents and educators, understanding this distinction is becoming an essential part of digital literacy.

Why does social media keep recommending similar videos?

Platforms such as TikTok, Instagram and YouTube use recommendation systems to predict what someone might want to watch next. They can consider signals such as viewing time, likes, searches and previous interactions, although the precise systems and settings differ.

The difficulty is that an action does not always reveal the reason behind it.

A teenager might watch a video about appearance because they enjoy the creator’s advice. They might also watch because something said at school has made them insecure, because they doubt the information, or because they are trying to understand a subject that worries them.

Those motivations are different, but their observable behaviour may look similar to the recommendation system.

The result can be a feedback loop: the platform recommends more content on the subject, the person encounters and watches some of it, and those interactions provide new signals for further recommendations.

Sometimes that produces a wonderful learning journey. An interest in paper airplanes might lead to aerodynamics, engineering and flight. At other times, the recommendations become increasingly repetitive, offering less variety and fewer opportunities to move away from an unwanted subject.

The problem is not necessarily one video. It is what happens when similar videos keep arriving.

What the European Commission found about TikTok

On 6 November 2026, the European Commission published preliminary findings concerning TikTok’s For You recommendation system.

The Commission raised concerns that repeated exposure to content involving issues such as body image, isolation and mental health could create harmful recommendation rabbit holes. It also provisionally found that TikTok’s existing controls may be insufficient to interrupt those patterns.

One particularly important concern was that viewing time could carry more weight than a person’s explicit request to see less of certain content through the Not interested function.

These findings are preliminary. TikTok has the right to respond, and the Commission has not issued a final infringement decision.

Nevertheless, the investigation exposes a question that matters to anyone using a personalised feed: if a platform offers a way to reject a recommendation, how do we know that it genuinely respects that choice?

In our full EU in Practice article, Why the Feed Won’t Let Go, we examine the independent research on recommendation rabbit holes, children’s reported experiences online and the implications of the European Commission’s findings.

We also explore the proposed EU KIDS Act, including its ambition to make online services safer by design—and the difficult balance between protecting children and respecting their privacy and growing independence.

Three things worth noticing in a child’s feed

Before searching for another parental-control setting, it helps to understand the experience itself.

1. Notice repetition, not just individual videos

A single video about dieting, a frightening news story or an upsetting social situation may not tell us much. A sequence dominated by similar material tells us something different about the environment the platform is creating.

Rather than immediately asking why a child watched a particular video, try asking whether they have noticed the same subject appearing more often lately.

This changes the conversation. The child is no longer being asked to defend every viewing choice. They are being invited to investigate how the feed behaves.

2. Separate what the child chose from what was recommended

Ask whether the child actively searched for a subject, followed a creator or simply kept encountering the material in their recommendations.

These distinctions matter because a personalised feed is not a neutral record of someone’s interests. It is a selection made by a system that uses behaviour to predict future behaviour.

A child may understand perfectly well that algorithms exist without realising how much those algorithms influence which subjects become prominent in their daily experience.

3. Check whether changing direction actually works

Most major platforms offer ways to influence recommendations. Depending on the service, age, location and app version, these may include Not interested, topic filters, feed resets or alternative views of content.

But different controls have different purposes. Pressing Not interested on one video does not necessarily remove a subject altogether. Resetting recommendations may give the system a fresh starting point, but it can begin learning again from subsequent behaviour.

The useful question is whether the unwanted pattern actually decreases after a change.

If content is distressing, there is no reason to keep watching it simply to test whether the algorithm responds. Stop the activity and prioritise the person’s wellbeing.

A two-minute activity you can try without opening an app

Here’s a way to introduce the idea to children without examining their accounts or exposing them to potentially upsetting content.

Imagine someone watches a 42-second video about folding a paper airplane.

What might the algorithm conclude? Perhaps that the person likes crafts, wants to learn about flight or enjoys paper-airplane competitions. But perhaps they are simply completing a school assignment.

Now imagine two possible feeds. One introduces how wings work, small experiments and the science of aviation. The other recommends increasingly similar folding challenges until nearly every suggestion concerns paper airplanes.

Neither journey is inherently harmful. But they demonstrate how recommendations can expand someone’s interests or keep returning to a narrow interpretation of them.

Ask your child which journey they would prefer, what the algorithm might have misunderstood and how they would tell the system they wanted something different.

That is the beginning of a useful conversation about digital choice—without making the child feel that their online behaviour is under interrogation.

Turn the conversation into a family investigation

We developed The Feed Detective, a free REDefine learning experience for families with children aged 9–17.

Through five short animated episodes and connected interactive activities, families follow Jamie’s paper-airplane story, explore how algorithms make assumptions, practise responding to unwanted recommendations and create a family digital agreement built around trust and shared responsibility.

An illustrated printable guide and activity sheets are also available. The activities require no registration, technical expertise or access to a child’s private social-media account.

Why the responsibility cannot rest entirely with parents

Parents can help children recognise uncomfortable patterns, use available controls and discuss what they encounter online. These are meaningful things to do, particularly when children feel able to ask for help without automatically fearing punishment.

But families cannot redesign commercial recommendation systems, examine internal platform risk assessments or establish whether a safeguard works across millions of accounts.

That responsibility belongs to service providers and the institutions responsible for overseeing them.

The proposed EU KIDS Act, introduced in September 2026, reflects this wider challenge. Alongside age-based access rules, it seeks stronger safety-by-design requirements, including changes to addictive design features and recommendation systems used by minors. The proposal is not yet law.

Its promise is that children should encounter safer systems, rather than being expected to navigate every foreseeable risk on their own. Its implementation will also need to protect children’s rights to privacy, information and participation.

These are the questions we examine in greater depth in our EU in Practice analysis.

Understanding the feed is something families can begin doing today. Making the systems themselves safer requires much more than another conversation at home.

Continue with the evidence and European policy: Why the Feed Won’t Let Go — Read on Substack

Make it practical with your family: Explore The Feed Detective — Free Interactive and Printable Toolkit


Frequently asked questions

Why does TikTok keep showing similar videos?

TikTok’s recommendation system uses viewing behaviour and other interactions to predict what someone might watch next. Repeated engagement with a subject can lead to more recommendations about that subject, even when the person’s original reason for watching was curiosity, concern or disagreement.

Does pressing Not interested stop unwanted recommendations?

Not necessarily. Not interested communicates a preference, but its effect depends on the platform’s recommendation system. If similar material continues to appear, other available controls or feed modes may be more appropriate.

How can parents teach children about social media algorithms?

Start with neutral examples, such as a hobby video, and compare different possible recommendation journeys. Ask what the system can observe, what it might misunderstand and how a person could change direction. REDefine’s free Feed Detective toolkit provides guided activities for doing this together.

Is the EU KIDS Act already in force?

No. The European Commission proposed the EU KIDS Act in September 2026. It aims to strengthen children’s online safety through age-related access rules and safety-by-design obligations, but the proposal must complete the EU legislative process before becoming law.