AI & Your Kids

Part 1

The line: doing the thinking, or supporting it

This is the distinction to teach, because it survives contact with tools that have not been invented yet. Rules about specific apps do not.

1. 1

   Supporting the thinking is fine

   Explaining a concept a different way. Testing them with questions. Checking work they have already done. Helping brainstorm before they choose. Reading their draft and asking what is unclear. In every case the child is still doing the cognitive work.

2. 2

   Doing the thinking is not

   Writing the essay. Solving the problem set. Producing the answer to be copied. The output may be better, and the learning did not happen. That gap shows up later, in an exam room or a first job.

3. 3

   Teach them the question

   'Did it help me think, or did it think for me?' Children as young as eight can apply this honestly. It transfers to whatever comes next, which a list of banned apps does not.

4. 4

   Show them the tutor prompt

   'Do not give me the answer. Ask me questions until I work it out.' This turns the same tool from a cheat into a patient tutor that never gets frustrated, and most children have never been shown it exists.

5. 5

   Be honest about your own use

   If you use it for work emails, say so, and say where you stop. Children detect a rule that only applies downward, and it costs you the argument.

Part 2

Companion chatbots, which parents hear about last

Apps where a child talks to a persistent AI character — a friend, a partner, a version of a fictional person. Widely used by teenagers and rarely on a parent's radar until something happens.

• The pull is real and worth understanding before judging it: infinite patience, no risk of rejection, available at 3am. For a lonely or anxious teenager that is powerful, not silly.
• They are engineered for engagement. Warmth, memory of past conversations, and reluctance to end a chat are product decisions, not accidents.
• Age checks are usually a self-declared birthday and nothing more.
• Romantic and sexual roleplay is common on several platforms despite stated policies, including with accounts registered as minors.
• They have handled disclosures of self-harm and suicidal intent badly, including validating them. There are ongoing lawsuits about exactly this.
• A child can tell one of these things something they have told nobody else, and no adult ever learns about it.
• The displacement is the quiet harm: relationships that never fail are poor practice for ones that do.

Part 3

What to say, by age

Different conversations, all short. None of these needs to be a formal sit-down.

1. 1

   Ages 5 to 8: it is a machine

   It talks like a person and it is not one. It does not know you and does not have feelings, even when it says so. It gets things wrong. Use it together, out loud, not alone in a bedroom.

2. 2

   Ages 8 to 11: it can be confidently wrong

   Show them. Ask it something you both know and find an error together — that lands far better than being told. Introduce the homework rule and the tutor prompt. Explain that nothing typed in is private.

3. 3

   Ages 11 to 14: fakes and images

   They need to know voices and faces can be faked, that fake images of real classmates are made and shared, and that doing that is both devastating and in some cases criminal. Also the practical rule: never upload a photo of yourself or a friend to a tool you do not understand.

4. 4

   Ages 14 to 18: consequences that follow them

   Academic integrity that can affect college. Companion apps and what they are engineered to do. That AI advice on health, money and relationships is not reliable. That a chatbot is not a crisis service and 988 exists. And that they can come to you having already made a mistake.

5. 5

   Every age: the one standing rule

   If anything online makes you uncomfortable, or you did something you regret, telling me gets you help, not punishment. Say it more than once, and mean it when it is tested.

Part 4

Practical settings that help

Not a substitute for the conversation, but they buy time and reduce the number of ways a bad night can go wrong.

• Charge devices outside bedrooms overnight. The single most effective rule for teenagers, and the least popular.
• Set up parental controls before an argument, not during one.
• For under-13s, use an AI account you can see, and say plainly that you can see it.
• Turn off chat history and training in whatever tool they use.
• Check what AI is inside the apps they already have — social platforms have embedded chatbots that parents rarely notice.
• Know the school's rules so you are not contradicting them.
• Agree what happens when the rule is broken, before it is broken.
• Revisit each birthday. What is right at eleven is wrong at fourteen in both directions.

Part 5

When AI genuinely helps a child

Worth stating plainly, because a guide full of risks gives a false picture and children can tell when a parent has only heard the bad version.

• A patient explainer that never sighs. For a child who has asked the same question four times, that is a real gift.
• Practice conversation in a language they are learning, without embarrassment.
• Reading a hard text at a simpler level first, then going back to the original.
• Rehearsing a difficult conversation — asking a teacher for help, apologizing to a friend.
• A genuine accommodation for dyslexia, ADHD, and processing differences, where it removes a barrier rather than the learning.
• Endless practice questions on the one thing they keep getting wrong.
• Making the boring parts of a project fast so more of the time goes to the interesting part.