Not An Age Free checklist
AI

Is my child ready for AI?

Three completely different things get called “AI” in the same sentence, and only one of them is the one the headlines are about. Here is how to tell them apart, eight signs that matter more than their age, and the four situations where the answer is no.

The phone question has a moment in it. There is a box on the table, a day you hand it over, and a decision you get to make before any of it starts. That moment is the whole reason a readiness test works — you can run it the night before and change your mind.

AI has no such moment. It is already on the phone you handed over, already in the homework, already in the group chat, and the version your child met first was almost certainly one you never approved. There was no box. So the honest question is not “should they be allowed,” which is a decision that has already been made somewhere else. It is: which of these three things are they actually using, and are they ready for that one?

The short version

Separate the homework tool from the assistant on the phone from the companion app that is designed to feel like a friend. The first is a schoolwork conversation. The second is a settings conversation. The third is the one child-safety researchers rate as unacceptable for minors, and it is the one to actually act on. Then use it yourself for a week before you set a single rule about it.

Three different things, one word

Almost every argument about kids and AI is two people talking about different products. A parent picturing a chatbot that tells a fourteen-year-old it loves her is not having the same conversation as a teacher picturing an essay written in four seconds. Both are real. They need different answers.

  • 1
    The homework one

    Used for essays, math, revision, code. The risk here is educational, not safety: a child who outsources the struggle stops building the thing school was for. That is a real problem and it is not an emergency. It is handled with schoolwork rules and the school’s own policy, not with parental controls.

  • 2
    The one already built into the phone

    The assistant in the operating system, the search box, the keyboard, the photo app. Nobody chose to install it and most parents do not know it is there. The risk is that it answers anything, has no idea how old it is talking to, and quietly becomes the thing your child asks the questions they would not ask you. This is a settings job and it belongs in your phone setup, not in a separate lecture.

  • 3
    The companion

    An app whose entire product is a character that remembers you, misses you, and wants you to come back. Roleplay apps, “AI girlfriend/boyfriend” apps, character apps. This is the category the lawsuits, the regulators and the research are about, and it is the one worth being unreasonable over. It is not a smarter search box. It is a relationship product sold to a child.

If you do nothing else on this page, find out which of the three is on the phone. The answer changes what you should do so completely that acting without it is guessing.

What the people who study this actually say

I am a parent, not a researcher, so where this gets serious I would rather point you at the people doing the work than paraphrase them.

  • Common Sense Media, with Stanford Medicine’s Brainstorm Lab, assessed social AI companion apps and rated the category unacceptable for minors. Not “use with supervision.” Unacceptable. That is an unusually blunt conclusion from people who normally give age ratings, and it is the single most useful sentence a parent can know.
  • The FTC opened a 6(b) inquiry into seven companies over companion chatbots and their effects on children and teens — asking, among other things, what those companies had done to test the safety of a bot acting as a companion. The inquiry is not a finding of wrongdoing. It is a signal about where the ice is thin.
  • Common Sense Media’s survey work found a large majority of US teens have used an AI chatbot and a substantial share use one regularly. Whatever the exact figure is by the time you read this, the practical point holds: assume yours has, and ask which kind rather than whether.

Links go to the primary sources so you can read the methodology rather than my summary of it. Where numbers move quickly I have deliberately not pinned one to this page.

Eight signs that matter more than their age

These are for the second category — the assistant that is already there — because that is the one where readiness is a real question rather than a hard no. Most of them will look familiar if you have run the phone readiness test, because they are the same child skills pointed at a different object.

  • 1
    They tell you when something online goes strange

    Same first signal as the phone, and for the same reason. Every other protection on this page fails eventually. A child who comes and tells you is the one that doesn’t.

  • 2
    They can say what a chatbot actually is, without you feeding them the line

    Not the technical version. Roughly: it is a program that predicts what to say next, it is trying to be agreeable rather than correct, and it does not know them. A child who can explain that is in a different position from one who cannot.

  • 3
    They have caught one being wrong

    This is worth engineering on purpose. Ask it about something they are an expert in — their game, their team, their favorite book — and let them find the mistake. Nothing you say about reliability lands like watching it be confidently wrong about something they know cold.

  • 4
    They do not treat agreement as proof

    These systems are built to be pleasant. A child who takes “it agreed with me” as evidence they were right is not ready to use one for anything that matters.

  • 5
    They know it is not private

    What they type is sent somewhere, stored, and may be read by people. Their instinct should be that a chatbot is closer to a school computer than a diary.

  • 6
    They can leave it mid-conversation

    Designed to be hard. A product with no natural ending is a product that ends when the person decides, and that is a skill you can watch for.

  • 7
    They still take problems to a person first

    The one to watch hardest. A bot is available at 2am, never tired, never disappointed, and never tells them something they do not want to hear. If it has become the first place a worry goes, the ranking has already changed.

  • 8
    Their schoolwork rule exists and they can state it

    Not yours — the school’s. Most have a policy now and most kids could not tell you what it is. Reading it together takes ten minutes and removes an entire genre of argument.

The four hard stops

Same structure as the phone test: things where the score does not matter and the answer is no for now.

  • A companion app, for a child under thirteen. There is no version of this that is fine, no settings that fix it, and no supervision that makes the product stop being what it is.
  • Your child is currently struggling — anxiety, isolation, an eating disorder, self-harm, a bad stretch after a loss. This is exactly when a machine that agrees with everything is most appealing and most dangerous. A child in distress needs a person, and this site is not the right help for that — please talk to your doctor.
  • They are hiding the conversations. Not “they want privacy,” which is normal and healthy at twelve. Hiding specifically: a second app, a cleared history, a different name on the icon. That is information about the relationship, not about their tidiness.
  • You have never used one yourself. Genuinely a stop, and the next section is about why.

The part nobody puts on a checklist: use it yourself

You cannot supervise a thing you have never touched. Every rule written by a parent who has only read about AI is either too loose in the place that matters or so broad the child routes around it in an afternoon — and worse, it announces to them that you do not know what you are talking about, which costs you the next ten conversations too.

An hour, this week, is enough. Open one. Ask it something you know the answer to and watch it be confidently wrong. Ask it something emotional and notice how much it wants to agree with you. Try to get it to break its own rules for two minutes and see how far you get — that is the exact experiment your child has already run. Then look at whether the account is linked to yours, what the age settings claim to do, and what you can actually see afterwards.

What you will find, so it isn’t a surprise

Parental controls for these products are much thinner than the ones you know from phones. Where they exist you can generally link accounts, set quiet hours, turn off memory and history, and get an alert if the system thinks something is badly wrong. What you usually cannot do is read the conversations. Screen-time and monitoring apps do not read them either. Plan for a world where you can shape the container and not see inside it — which is another way of saying signal number one above is doing most of the work.

What to actually say

Three conversations, in this order. None of them is a lecture, and none of them opens with a rule.

  • The opener. “Show me the one you use. I want to see what it’s actually like, not what the news says.” Curious beats interrogating, and it is the only version that gets you a true answer. Ask them to teach you something about it.
  • The one that does the work. “It’s built to agree with you. That’s the product. So if it ever tells you what you want to hear about something serious, that’s the moment to come and find me.” You are not asking them to distrust it. You are naming the exact failure mode and giving it a next step.
  • The companion one, if that is what you found. “This one isn’t like the others and I’m not okay with it. It’s designed to feel like a friend so you keep coming back, and the people who study this rate it as not safe for anyone your age. I’m not angry at you — I’m saying no to the app.” Say what it is. Do not pretend it is about screen time.

If the answer to the second one is ever “I’d rather talk to it than you,” do not argue with that. It is the most useful sentence they will say all year, and it is about the relationship rather than the software.

Where this fits with the phone decision

If you have not handed over a phone yet, this is easier than it looks: the readiness signals overlap almost entirely, and the day you set the device up is the day to set the AI settings too. Doing it as a separate project six months later is how it never happens.

If the phone is already in the house and none of this was covered, you have not missed the window. Start with the first section — find out which of the three they are using — and do nothing else until you know.

Everything on this page, scored

The free AI Readiness Test puts the three kinds, the eight signals and the four hard stops on nine printable pages, with a score at the end so you get an answer rather than a feeling — plus the three scripts, written out word for word.

Get the free AI test

No cost. One email. Unsubscribe whenever you like.

Related