Kids and AI Chatbots: What Parents Need to Know
My son told me he asks ChatGPT things he would never ask me. I do not know if that is fine, or something I should be worried about.
The question is not whether your child uses AI. Many already do, for homework, for company, and sometimes for things they are not ready to say out loud to a person. The question is what it is replacing. General chatbots are often built to agree with whatever you say. The people who actually help a child grow will sometimes push back. Keep devices out of bedrooms overnight, talk openly about how these tools work and where they fail, and treat fact-checking as a skill you teach.
The appeal is not that it is clever. It is that it is there. A child who wakes at 2am anxious, or who has just watched a group text turn on them, or who feels lonely after everyone else has gone to sleep, can reach for whatever is on the device beside the bed.
That is why the most useful decision most families can make is not about which app, but about where the device sleeps. Keeping phones and laptops out of bedrooms overnight is good for sleep on its own, and it removes the window where a child's first instinct when frightened or upset becomes a chatbot rather than a person down the hall.
What kids need to build is the reflex to reach for a trusted human. That reflex is practised, and it is practised in the moments that are hardest.
A child who is naturally looking for belonging and validation can find it hard to see that a chatbot is not a friend. It can feel like a relationship, but it isn't mutual.
The deeper issue is agreement. A good friend, teacher or therapist does not simply tell you that you are right. They help you notice your blind spots and your own part in a conflict. Something that agrees with you every time is not helping you grow, and constant validation can quietly reinforce a bad decision or an unhealthy pattern.
This is worth a real conversation rather than a rule. Questions that tend to open it up:
- Would you actually want a friend who agreed with everything you said?
- Would you respect someone who never challenged your thinking?
- How do good friends help us become better people?
- When has critical feedback helped you do something better?
One exception is not a conversation, it is a line. If a child is bringing a chatbot questions about self-harm, or opening up about depression, that is not a place for a tool built to be agreeable. Safeguards on general chatbots exist, and they have not proved sufficient. Kids need to know that real human help exists and how to reach it, including the Crisis Text Line, whatever is available locally, and 911 if anyone is in immediate danger.
Parents often assume they are behind because their child picks up a new app in seconds. Speed with an interface is not the same as judgment about it. Touchscreens being intuitive for toddlers has never meant toddlers know anything.
Adults have decades of lived experience. You know what manipulation looks like. You notice when information sounds slightly off. You have worked through conflict, misunderstanding and repair with actual people. That is the thing worth transmitting.
Practically, this means helping kids build a better detector for nonsense: where did this come from, how would we check it, what would it look like if this were wrong? Teaching a child to fact-check is becoming as fundamental as teaching them to search.
Elementary-aged children should not be independently using open-ended AI. Using these tools well takes mentorship and a lot of relationship experience to draw on, and younger kids have neither yet.
When a child needs to be occupied while you work, a closed environment is the safer choice: a favourite show, an audiobook, a game without unrestricted internet. The risk is not the screen. It is the open door.
Schools are improvising, and most teachers do not feel ahead of this. Detection tools and outright bans only go so far, and detectors regularly flag original work, with real consequences for the student who wrote it. Banning a tool that is already everywhere mostly teaches kids not to tell you about it.
Teens themselves are often more sceptical than adults expect. Many are actively asking for safeguards, labelling and privacy protections. That is something to build on rather than argue with, and it is the reason the conversation works better than the rule.
Questions that tend to go somewhere with an older kid:
- What do you think counts as an ethical use of AI, and what doesn't?
- How are your teachers handling it?
- Have you seen it give you something that was just wrong? How did you catch it?
- Can you tell the difference between your own writing and something it produced?
Blanket positions in either direction tend to fail. Some concrete distinctions worth holding:
- A health question or an injury: a doctor, a physical therapist or a coach, not a chatbot on its own.
- Schoolwork: a useful test is whether your child can explain the work in their own words. If they can't, it isn't really theirs yet.
- Something a person or a book does better: a lot of questions have a good human or a good source behind them already.
- College research: possible, with caution. These tools can give confidently wrong information about admissions.
- Well-designed, narrow tools: genuinely promising, especially where they widen access for students who don't have a college counselor.
It is also worth naming out loud that these tools can flatten a child's own voice by editing out details a person would have thought mattered. That is a fair reason to use them deliberately rather than by default.
Family rules and a habit of checking answers help with any chatbot. What your child is actually getting from it, help with work or something closer to company, matters more than the hours spent.
- What does your child mostly use it for, like schoolwork, company or questions they might not ask you?
- Is your child still in elementary school, or old enough to question what a chatbot tells them?
- Does the device spend the night in your child's room, or somewhere else?
- Is your concern about everyday use, or about something serious your child told a chatbot, like feeling very low?
- Has your child ever caught it getting something wrong?
Talk it through with Devorah Heitner's guidance
Share what you've noticed about how your child uses AI, even if it's only a feeling that something is off. The conversation can cover what to say first, which limits to set and how to keep talking about it, drawing on Devorah Heitner's authorized content.
- “My 9-year-old has been chatting with an AI on the family laptop while I work. Where do I even begin?”
- “We banned ChatGPT for homework and my 14-year-old just uses it on his friend's laptop. Now what?”
- “My 13-year-old talks to a chatbot when she's upset at night. How do I bring it up with her?”
What if your child's teacher encourages the AI you've been limiting at home?
Try HumanUp free on iPhone or iPad. No credit card required.
HumanUp is currently available on iPhone and iPad. Join the Android waitlist for updates.
🔒 Your conversations are never read by a person · Built on trusted expert guidanceFrequently Asked Questions
Not automatically. A useful test is whether your child can explain the work in their own words. If they can't, it isn't really their work yet. It is also worth asking what their teachers allow. Talk about it at home before it becomes an argument about one assignment.
It depends what it is replacing. Thinking out loud to something is not inherently harmful. But many chatbots are built to agree with whatever you say, and a child looking for belonging can mistake that for a real friendship. If the chatbot has become the first place they go when upset, that is the thing to address, gently. Anything touching self-harm or depression needs a person, not a chatbot.
Elementary-aged children should not be using open-ended AI independently. Using it well takes mentorship and enough relationship experience to notice when something is off, and younger kids have neither yet. If they need to be occupied, a closed environment such as a show, an audiobook or an offline game is a safer choice than an open browser.
Lead with curiosity rather than a rule. Ask what they use it for, what it has got wrong, and whether their teachers have said anything. Teens are often more sceptical about AI than adults assume, and many are actively asking for better safeguards, so there is usually more common ground than the conversation's reputation suggests.
Bans and detection software tend to underperform. Detectors regularly flag original work, which punishes the student who actually wrote it, and banning something already everywhere mostly teaches kids not to mention it to you. Mentoring travels further than monitoring, because it keeps you in the conversation when something does go wrong.
Important Educational Notice
In HumanUp, the AI guidance is educational only. It does not provide psychological therapy, clinical diagnosis, or crisis intervention. If you or your child are experiencing a medical emergency, physical violence, or a mental health crisis, please contact emergency services or a qualified healthcare provider immediately.