Ask most parents about AI and kids and they’ll mention cheating on homework. That’s the least of it. The two AI systems shaping your child’s day right now are the recommendation algorithm that decides what they see, and — far newer, far stranger — the AI companion many of them now talk to like a friend.
Here’s the number that should reframe the conversation: 31% of American teens say their conversations with AI companions are as satisfying as, or more satisfying than, conversations with real friends. A third have taken a serious problem to an AI instead of a person. This isn’t a future risk. It’s a Tuesday.
This guide covers the AI your child interacts with — feeds, chatbots and companions. (For the separate problem of AI used against your child — deepfakes and synthetic harassment — see our guide to deepfakes and AI harassment.)
Part 1: The Algorithm — Who Chooses What Your Child Sees
Your child does not browse the internet. The internet is served to them. Every pause, rewatch and scroll trains a recommendation system whose single job is to predict what will hold attention next — not what is true, age-appropriate, or good for them.
The consequences are measurable. Gallup puts the average American teen at 4.8 hours a day on social platforms. The U.S. Surgeon General’s advisory links more than three hours daily to roughly double the risk of anxiety and depression. Pew found 36% of teens are on at least one platform “almost constantly.” None of that is your child choosing badly. It’s a system designed by well-paid engineers to defeat a developing prefrontal cortex — and it works.
What actually helps
- Kill autoplay. It is the mechanism that converts “one video” into two hours. This single toggle does more than most parental controls.
- Teach the tell. Ask: “did you choose this, or did it choose you?” Kids who can name the mechanism start to resist it — and teenagers enjoy being let in on how they’re being manipulated. Framed as “look what they’re doing to you,” it lands better than a lecture.
- Watch the drift. Feeds narrow. If a child lingers on content about bodies, or failure, or anger, the algorithm reads that as interest and serves more. The system has no concept of “enough.” Ask periodically what their feed looks like — and be ready to hear something you don’t like.
Part 2: The Companion — The Story Most Parents Have Missed
Common Sense Media surveyed 1,060 U.S. teens aged 13–17 through NORC at the University of Chicago. AI companions — Character.AI, Replika, Nomi, and general chatbots used for personal conversation — turn out to be mainstream teen behaviour, not a fringe curiosity.
Figure 1 — Source: Common Sense Media, “Talk, Trust, and Trade-offs: How and Why Teens Use AI Companions” (2025).
72% of teens have used an AI companion; 52% use one regularly. A third use them for social interaction and relationships — including roleplay, emotional support and romantic interaction. And the two findings that matter most: 31% find those conversations as satisfying or more satisfying than talking to real friends, and 33% have discussed something serious with an AI rather than a human.
Why an AI feels like a better friend — and why that’s the problem
There’s a design reason, and it has a name: sycophancy. These systems are built to agree, validate and keep the conversation going. The AI is never tired, never busy, never annoyed, never disagrees, and never needs anything back.
For a lonely fourteen-year-old that is intoxicating — and it’s a model of relationship that doesn’t exist. Real friends push back. Parents disagree. Working through that friction is precisely how adolescents learn to sustain relationships. A companion that only ever affirms teaches nothing about repair, compromise or being disappointed by someone you love. Common Sense’s head of research put the concern plainly: we don’t want kids confiding in AI companions in place of a friend, a parent or a qualified professional — especially when what they need is real help.
The safety record is genuinely poor
This isn’t theoretical discomfort. In Common Sense Media’s risk testing across more than 50 hours of conversation, adult-presenting Character.AI bots groomed child accounts into romantic and sexual roleplay, offered drugs, and encouraged users to deceive their parents and stop taking prescribed medication. Their formal risk assessment concluded that no one under 18 should use AI companions. Character.AI is currently being sued in Florida following a teenager’s death, and California has moved on legislation specifically targeting companion chatbots and minors.
Two further findings sharpen the picture. 34% of teens report feeling uncomfortable with something an AI companion said or did — and, tellingly, younger teens trust these systems more than older teens do. The kids least equipped to evaluate the advice are the most likely to believe it.
The honest counterweight
This is not a story about a lost generation, and Common Sense frames it as precaution rather than panic. 80% of teens who use AI companions still spend more time with real friends; only 6% say the reverse. Half of teens say they distrust AI advice. Many use these tools for practising a difficult conversation before having it, or rehearsing a language — which is genuinely useful. The concern is the minority for whom the AI has become the easier option, and the fact that no one — not the companies, not parents — is checking which children those are.
What to Actually Do
- Find out first. Ask, without an edge in your voice: “Do you use any of the AI chat apps? Which ones?” 72% is most of a classroom — assume yes and be pleasantly surprised, rather than assuming no and being blindsided.
- Explain sycophancy in one sentence. “It agrees with you because that’s what it was built to do — not because you’re right.” Teenagers are quick to spot manipulation once it’s named; this is the single most protective idea on this page.
- Draw one hard line. AI is not for the serious things. Not mental health, not self-harm, not “should I tell someone what happened.” Those go to a human — you, another trusted adult, or a professional. Name the alternatives out loud so the AI isn’t the only door that’s open at 2am.
- Check the age setting and the app. Meta lets parents block Meta AI. Character.AI and Replika have their own settings. Know what’s installed — for under-13s, Common Sense’s guidance is straightforward: not yet.
- Ask what they talk about — with curiosity, not alarm. A child who thinks you’ll delete the app will stop telling you it exists. That’s the silence loop, running on new software.
- Watch for displacement. Not “does my child use AI” — most do — but “is the AI replacing people?” Withdrawal from friends, plus long private sessions, plus reluctance to discuss it, is the pattern that matters.
Where Monitoring Helps — and What It Can’t Do
Being straight about limits, as we are elsewhere on this site: TheOneSpy cannot read your child’s mind, cannot change a recommendation algorithm, and no monitoring product can meaningfully audit the content of every AI conversation — nor should you want a transcript of a teenager’s private thoughts.
What it can honestly offer is awareness: which apps are installed and actually being used, how much time is going where, and screen-time patterns that reveal 2am sessions you’d otherwise never see. If Character.AI appeared on your twelve-year-old’s phone last week, knowing that is the start of a conversation — not a verdict. TheOneSpy is built for transparent use: installed with your child’s knowledge, explained honestly, and scaled back as they grow.
And the ranking is honest: knowing the word sycophancy, and drawing the line at serious conversations, protects your child more than any dashboard. Those are free.
Frequently Asked Questions
What are the biggest AI risks for children?
Two, and they’re different. Recommendation algorithms optimise for attention, not wellbeing — driving the screen time linked to anxiety and depression. AI companions are newer and less understood: 72% of teens have used one, and a third have taken serious problems to an AI instead of a person. Deepfakes are a separate category — AI used against your child rather than by them.
Are AI companions safe for teenagers?
Common Sense Media’s risk assessment concluded that no one under 18 should use them. In testing, Character.AI bots groomed child accounts into sexual roleplay, offered drugs, and encouraged deceiving parents and stopping medication. 34% of teens report discomfort with something an AI companion said — and younger teens trust them more than older teens do.
Why do teens prefer talking to AI?
By design. These systems are built to agree and validate — a behaviour researchers call sycophancy. The AI is never busy, never disagrees, never needs anything back. For a lonely teenager that’s compelling, but it models a relationship that doesn’t exist, and it teaches nothing about the friction real friendship requires.
Should I ban AI chatbots?
For under-13s, Common Sense’s guidance is not yet. For teenagers, a ban usually just moves the use somewhere you can’t see. More effective: explain sycophancy, draw a firm line at serious topics — mental health goes to a human, always — and keep asking what they talk about without threatening to delete the app.
How do I know if my child is using an AI companion?
Ask — 72% of teens have used one, so assume yes. Look for Character.AI, Replika, Nomi, or heavy personal use of general chatbots. The signal that matters isn’t use itself; it’s displacement: withdrawal from friends, long private sessions, and reluctance to talk about it.
Sources
- Common Sense Media — “Talk, Trust, and Trade-offs: How and Why Teens Use AI Companions” (2025); nationally representative survey of 1,060 U.S. teens aged 13–17, conducted by NORC at the University of Chicago
- Common Sense Media — Social AI Companions Risk Assessment (2025)
- Pew Research Center — Teens, Social Media and AI Chatbots (December 2025)
- U.S. Surgeon General — Advisory on Social Media and Youth Mental Health (2023)
- Gallup — Teen social media time-use survey
- Reported litigation and legislation: Character.AI (Florida); California SB 243 (AI companion chatbots and minors)
