There is a particular kind of parenting advice that is designed to produce anxiety rather than action — advice that catalogues every danger in exhaustive detail while leaving the parent with no clearer sense of what to do than before they started. The previous eight pieces in this series have, of necessity, spent considerable time with the dangers. This one is about what to do.
The evidence on what actually protects children in AI-saturated digital environments is more specific than most parents realize, and more complicated than the parental control industry — a growing sector that benefits from parental fear — would prefer. Not all interventions work. Some that are marketed as protective may be counterproductive. And the single most protective thing a parent can do does not require software at all.
What the Research Shows About Monitoring
The two most widely used AI-powered monitoring tools for families — Bark and Qustodio — take fundamentally different approaches, and the difference matters.
Bark uses AI to scan messages, images, and audio across more than 30 platforms for 29 categories of potential harm — cyberbullying, self-harm content, depression signals, drug references, sexual content. When it detects something concerning, it alerts parents rather than sharing the content itself. As of 2024, the platform claimed to protect nearly 7 million children and reported detecting approximately 4 million cyberbullying incidents and 2 million self-harm incidents. Its model is notification rather than surveillance: the parent learns that something is wrong without reading every message.
Qustodio offers more granular controls: content blocking, location tracking with geofencing, detailed activity logs, time limits by app. It is better suited to younger children who need firm external limits.
The research on monitoring tools contains a finding that is easy to dismiss but important to take seriously: hidden monitoring is counterproductive. Studies consistently find that children who know they are being monitored without being told — who discover surveillance that was not disclosed — are less likely to talk to parents about online problems, more likely to find workarounds, and less likely to internalize the values behind the rules rather than simply the rules themselves. "Parents cannot discuss online experiences with children if they don't know they're being monitored," one researcher summarized, "and without reflective conversation, children are less likely to learn from the experience."
Monitoring used transparently, as a tool that supports conversation rather than replacing it, has a different evidence base. The question worth asking before installing any monitoring software is not "will this let me see what my child is doing?" but "will this make it more or less likely that my child will tell me when something goes wrong?"
What Works: The Evidence Base
A meta-analysis of school-based digital wellness interventions identified the characteristics that consistently predicted effectiveness: led by trained external providers, actively involving parents, targeting at-risk populations rather than general populations, and using therapy-based approaches rather than information-only presentations.
For family-level interventions, the most reliably supported practices are simpler than most parents expect.
No devices in bedrooms overnight. This is among the most robustly supported findings in the research. Bedroom access to devices is associated with sleep disruption, reduced in-person socialization, and increased exposure to harmful content — all of which compound. A bedroom device ban is not technically complex. It is socially and emotionally complex, particularly for teenagers who experience significant peer pressure around device access. But the evidence for its protective effect is consistent.
Conversation over content filtering. Research finds that children in families where technology is openly discussed — where parents know what platforms their children use, have seen them in use, and talk about what they encounter — are more resilient to digital harms than children in families that rely primarily on technical controls. The mechanism is what developmental researchers call authoritative parenting: setting clear expectations while maintaining a relationship in which the child feels comfortable raising problems. Children who can tell their parents about a frightening encounter online are protected by that disclosure. Children who fear disclosure — because they fear punishment, device loss, or parental overreaction — face the consequences of that encounter alone.
Be specific about the offer. Research on why sextortion victims stay silent consistently identifies fear of device loss as the primary barrier to disclosure. A parent who has explicitly said — before anything has happened — "if something frightening or embarrassing happens online, you can tell me, and I will not take your phone away" has created a different environment than one who has not. The specificity of the promise matters. It should be made clearly, without conditions, and repeated.
Age-appropriate delays. Surgeon General Vivek Murthy has called for no smartphones before high school and no social media before age 16. Jonathan Haidt's research suggests that delaying social media access, specifically, by even a few years during the most neurologically vulnerable period of early adolescence produces meaningful reductions in anxiety and depression. The practical argument against this — that a child will be socially isolated if they do not have what their peers have — is real and should not be dismissed. But the evidence suggests the cost of the delay is smaller, and the benefit larger, than most parents assume.
The Limits of Technical Controls
The parental control industry generates significant revenue from parental anxiety, and it is worth being clear-eyed about what technical controls can and cannot do.
Content filters can block access to known categories of harmful material. They cannot block material that has not been categorized, cannot prevent a child from accessing material on a friend's device, and cannot address the social dynamics — peer pressure, loneliness, the desire for belonging — that make harmful content appealing to at-risk children. They work best as a baseline for younger children who are not yet ready for unsupervised internet access, not as a substitute for conversation and relationship for older children.
Time limits function as expected: they reduce the time a child spends on a specific platform. They do not reduce the desire for that time, the social anxiety that time-limited access creates when peers are available, or the intensity of engagement within the limited window. Research on time limits finds modest effects on screen time and weak effects on the mental health outcomes that are the actual concern.
The American Academy of Pediatrics updated its guidance to reflect this complexity. Strict screen time limits — the "no more than two hours per day" guidance that parents of teenagers have always found unenforceable — have given way to a more nuanced framework: prioritize quality over quantity, co-view and discuss content with younger children, create phone-free spaces and times, and treat the family media plan as a living document rather than a compliance checklist.
The Family AI Policy
One practice that is gaining evidence support but remains underutilized is the explicit family conversation about AI specifically — distinct from the general screen time discussion that most families have had. Children who understand, at an age-appropriate level, how recommendation algorithms work, why AI-generated content exists, and what "synthetic" means in a media context, are better equipped to question what they encounter. They know to ask: is this real, and who made it, and why?
This does not require technical sophistication. It requires the willingness to sit with a child and say: the video you just watched was chosen for you by a machine that wanted you to keep watching. The machine doesn't know you. It knows patterns. And sometimes the patterns it follows are not good for you.
That conversation — repeated over years, updated as the technology evolves — is what the research calls media literacy, and it is among the most durable protections available to any family. In the final piece of this series, we examine what raising children with that kind of literacy looks like in practice — and what it means to raise children who can live well in a world that AI is still building.
This is Part 9 of "Raising Children in the Age of Intelligent Machines," a 10-part series from PeopleSafetyLab on the intersection of AI and family safety.