The algorithm behind your child’s feed has more behavioural data on your child than you do. It knows what they watch the longest. It knows what they re-watch. It knows what they pause on. It knows what they scroll past. And it uses that knowledge to build a feed that gets harder to put down with every interaction. Once you understand what the algorithm knows and why, you can start reclaiming that knowledge for yourself.
What the algorithm actually tracks
Every platform your child uses tracks dozens of behavioural signals on every video, post, or item shown. The signals are detailed, continuous, and stored permanently. The information is not hidden. It is openly described in the platforms’ developer documentation and in research papers from the companies themselves.
Among the things tracked on most major platforms:
- How long your child watched each video, to the millisecond
- Whether they re-watched it
- Whether they paused, and where
- Whether they scrolled past quickly or slowly
- Whether they liked, commented, shared, or saved
- What time of day they tend to engage most
- What they tend to engage with after specific other content
- Their location, device, time zone, and language
- Patterns across all of the above, weighted by recency
This is not surveillance for its own sake. Each signal is used to refine the model of what content will hold your child’s attention longest. The longer they stay, the more advertising revenue is earned. The system has a clear goal, and it gets better at achieving it the more your child uses the platform.
What the algorithm figures out
From these signals, the algorithm infers things about your child that are surprisingly specific. Their interests, fears, current obsessions, mood patterns, social anxieties, identity questions. Not as a list of facts, but as a probability distribution that shapes the next video shown.
Some of this would be obvious to a parent paying attention. A child interested in a particular video game will be served more videos about that game.
Other parts are subtle and sometimes invisible to parents. The algorithm picks up on things the child has not told anyone. A growing interest in a particular kind of body image content. A drift toward darker humour. An emerging concern about a specific social dynamic at school. The signals leak into the algorithm before they leak into family conversation.
Why this matters more than parents realise
Children form their identity through experience and reflection. The experiences they have shape who they become. For previous generations, those experiences happened mostly in the physical world: school, friends, family, the books they chose, the places they went. Parents had reasonable line of sight on the inputs.
For the current generation, a meaningful portion of the formative experience is happening inside an algorithmic feed that the parent cannot see and the child cannot fully describe. The content is personalised, ephemeral, and never seen by anyone else. The parent is not in the loop on the most influential daily input in their child’s life.
This is the part of modern parenting that has no precedent. There is no equivalent of “the algorithm” in any earlier generation of childhood. The closest thing is television, but television was the same for everyone in the household, and it was finite. The algorithm is unique to your child and infinite.
What “the algorithm knows your child better than you do” actually means
This phrase is provocative on purpose, but it is also literal in one specific sense. The algorithm has more granular, real-time, behavioural data on your child than you do. It updates the model continuously based on observed behaviour rather than assumed behaviour.
It does not mean the algorithm understands your child better than you do. You understand your child as a person. The algorithm understands your child as a behavioural data set. These are different kinds of knowledge.
The risk is that the algorithm’s knowledge shapes your child’s experience while your knowledge increasingly does not. Over a year, two years, five years, the influence compounds. Your child’s interests, sense of humour, worldview, and aspirations are increasingly shaped by what was fed to them by a system optimised for one outcome. More time on screen.
How to reclaim the knowledge
The work is not technical. It is relational.
1. Spend more time in conversation than the algorithm spends predicting
The algorithm gets its knowledge from observation. You can get yours from conversation, which is faster and richer. A 15-minute conversation about what your child found funny this week tells you more than the algorithm could observe in days.
2. Watch with them sometimes
Sit beside your child while they scroll, occasionally. Not to police it. To see what the algorithm has chosen for them. The pattern in the feed tells you what the system thinks your child cares about. Sometimes the picture matches what you would expect. Sometimes it surprises you, and the surprise is useful.
3. Reduce the share of life the algorithm sees
Every screen-free moment is a moment the algorithm cannot observe and therefore cannot influence. The car ride, the dinner table, the half-hour before bed. These are moments that develop your child’s identity in ways the algorithm cannot reach.
4. Reset the algorithm occasionally
Deleting and reinstalling a platform every few months wipes much of the personalised model. It does not undo the influence already exerted, but it does break the compounding. The first few days after a reset are noticeably less compelling.
The longer view
The algorithm is not going away. The systems will keep getting better at predicting your child. The only sustainable response is to make sure your understanding of your child grows alongside the algorithm’s understanding, and that the parts of life the algorithm cannot reach stay protected.
That is what reclaiming looks like at this layer of the work. Not switching off the algorithm. Just making sure it does not become the most influential force in your child’s developing identity.
The bigger picture of why this work matters is in the Let’s Get Them Back manifesto. The full set of tools sits in the Complete Guide to Screen Time.
Frequently Asked Questions
What does the algorithm know about my child?
A surprising amount. How long they watch each video, what they re-watch, what they pause on, what they engage with, their patterns across time of day, and dozens of other behavioural signals. The model updates continuously based on observation.
Can I see what the algorithm knows?
Most platforms do not give parents direct access to the model. The closest you can get is to look at the recommended feed your child is served. The pattern in the feed reflects what the system thinks your child cares about.
Is the algorithm dangerous for my child?
Not inherently. The risk is that it becomes the most influential daily input on your child’s developing identity, especially when the parent is not in the loop on what is being shown. Reducing exposure and staying in conversation are the main protective moves.
Can I reset the algorithm?
Mostly yes. Deleting and reinstalling a platform every few months wipes much of the personalised model. The first few days after a reset are noticeably less compelling, which is a useful signal of how much the model had been doing.
Do all platforms work this way?
The major social platforms all use algorithmic feeds with personalised recommendation. Email, calendar, and educational apps generally do not. The pattern matters in any platform built around content discovery rather than utility.
Will the algorithm change as my child grows?
It will continually update based on what they engage with. Children’s algorithms shift dramatically during puberty as the content they engage with shifts. This is a useful inflection point to do an algorithm reset.
Talk to Cleo
Get a plan that reclaims what the algorithm knows. Free, no sign up.
