Luminae

When AI becomes the first place your child asks for advice

By Sean Corbett

When your child asks an AI chatbot for advice, how do they decide whether to trust the answer? A tool that explains homework clearly can also answer questions about friendship, health or something they feel unable to discuss at home. Parents need to understand that change of role.

If you read nothing else

Ask your child what kinds of questions they take to AI, then look at an everyday answer together. Before acting on advice, ask: “What could happen if this is wrong?” Important claims need checking outside the conversation. Health worries and serious personal problems need a trusted adult or qualified professional involved. Make sure your child knows who they can turn to, including when they would rather speak to someone other than you.

Start with what your child asks AI

Homework is an obvious place to begin a conversation about AI. It may not be where that conversation needs to end.

Imagine a teenager asking the same chatbot why the Roman Empire collapsed and, a few minutes later, whether a friend is deliberately excluding them. These are illustrative examples, but the change in subject matters. A clear explanation of history gives us little reason to trust advice about a particular friendship.

Ofcom’s 2025/26 report found that 56% of UK children aged 8 to 17 had used AI. Among users, 51% had used it for schoolwork or homework. About one in ten had used it as someone to talk to or as a friend. Its smaller qualitative study also recorded children asking about health worries and relationships. Those interviews illustrate individual experiences; they do not establish how common each behaviour is.

A useful opening question is: “What kinds of things do you ask it?” If they are willing, ask them to show you an example they are comfortable sharing. Listen to what they find helpful before discussing what concerns you.

Why a chatbot answer can feel convincing

A conventional search result points towards other people’s work. A chatbot usually presents a composed answer in its own conversational voice. Many search engines now offer AI summaries too, while some chatbots provide links. The distinction is less tidy than it once was, but there is still a practical difference between reading an answer and examining the sources behind it.

A chatbot can rephrase an explanation, answer follow-up questions and respond without an awkward pause. That can help a child understand a difficult subject or put a question into words. It can also make a response feel more dependable than the evidence warrants.

In a preregistered experiment by Kim and colleagues, 284 adolescents aged 11 to 15 and their parents read two versions of a chatbot exchange about a social problem. One used language that suggested a closer personal relationship; the other stated clearer boundaries around its role as AI. Teenagers rated the more relational version as more trustworthy and emotionally close, although they judged the two versions similarly helpful. This was a response to written transcripts, not a test of prolonged chatbot use or proof of dependency.

That finding gives parents something specific to discuss. Ask which parts of an answer make it seem trustworthy. Is it the supporting evidence, or the reassuring way it is written?

Personal advice depends on missing context

Consider this invented example: “My best friend has ignored me all week and went out without inviting me. I think she’s deliberately excluding me. What should I do?”

A chatbot might suggest a calm conversation or offer another explanation. Either could be helpful. But it is responding to the account it has been given. It may be missing an argument from last week, a misunderstanding or something the teenager has found difficult to say.

Some systems can draw on earlier conversations or other information supplied to them. They may identify distress in a message and suggest seeking help. That does not give them a dependable understanding of the child’s circumstances, or make them a substitute for someone who can take responsibility and act.

Repeatedly useful answers may encourage a child to turn to the same system for increasingly personal decisions. That is a reason to discuss where its advice stops being enough. Ask: “What might it be missing here?” Then help your child identify someone who could understand more of the situation.

When health or wellbeing is involved

A chatbot may help explain an unfamiliar medical term or organise questions for an appointment. Do not rely on it to decide whether symptoms are serious, select medication or work out a dose. A plausible answer can still be wrong, and a general chatbot cannot examine your child.

The evidence about mental health support needs careful reading. A 2025 review by Feng and colleagues brought together 31 trials of chatbot interventions. It found benefits on some measures, but the eligible age range was 15 to 39 and the overall effectiveness of generative systems remained inconclusive. Those results do not establish that an everyday chatbot is a safe or effective source of mental health care for a child.

Young people are nevertheless using them for advice. A US survey published in JAMA Pediatrics in 2026 found that 19.2% of respondents aged 12 to 21 had used a chatbot for mental health advice. Among those users, 63.3% had told nobody. The survey measured reported use and opinions, not whether the advice improved health. It also included adults and should not be read as an estimate for British children.

If your child is struggling, involve a trusted adult and an appropriate health professional. You do not need to wait until a problem becomes urgent. If chatbot use is displacing sleep, usual activities or relationships, ask what is happening and seek support. These changes deserve attention whatever their cause.

In England, NHS 111 can direct you to help when medical attention is needed now and you are unsure what to do. That page also links to urgent-care information for Scotland, Wales and Northern Ireland. Call 999 for a life-threatening emergency. Readers elsewhere should use their local healthcare and emergency services.

Help your child check an answer

Choose an ordinary question you can investigate together. Ask the chatbot for an explanation and look at one factual claim in its reply. Follow any cited link and check whether the source actually supports the claim. If the answer has no source, find a reliable one independently. For schoolwork, that might be the material their teacher has provided.

Asking the chatbot “Are you sure?” can prompt a revision, but it is not an independent check. Nor does agreement between two chatbots establish that an answer is correct.

The amount of checking should reflect the consequences. An imperfect suggestion for a story title is easily discarded. Advice affecting health, money or an important relationship needs more scrutiny and, where appropriate, a person qualified to help.

Try a short conversation such as: “I can see why that answer helped. Before you act on it, shall we check the part that matters most?” The aim is to practise a habit your child can use without you sitting beside them.

Agree boundaries that fit your child

These are practical suggestions for applying the evidence, rather than a tested programme. Adapt them to your child’s age, needs and growing independence.

  • Check the service’s current age requirements and parental permissions before use. A tool being available on a device does not mean it is appropriate for every child.
  • For younger children, work through permitted uses together. With older teenagers, agree which decisions need an outside check and respect their need for space to ask ordinary questions.
  • Explain that a conversation which feels private is not necessarily confidential. Avoid entering identifying details, medical records or other people’s private messages. Check the service’s privacy settings together.
  • Agree what your child will do if a reply is upsetting, confusing or encourages something unsafe. They can stop the conversation and speak to a trusted adult. They do not owe a chatbot a response.

The NSPCC’s guidance on talking to children about AI offers further help with these conversations. Parental settings can support the boundaries you agree, but they cannot establish whether every answer deserves trust.

Make it easier to ask a person

Some questions are embarrassing. A teenager may find it easier to type them than to watch a parent’s reaction. If your child tells you they have asked AI about something personal, start with the concern that led them there.

You might say: “Thank you for telling me. What was worrying you, and did the answer help or make things harder?” You can discuss the chatbot once you understand the problem.

Agree who else they could approach if they do not want to speak to you. Depending on the concern, that might be another trusted adult, someone in the school’s pastoral team or a health professional.

Parents do not have to match a chatbot’s speed or range of answers. What matters is that a child can bring an uncertain or uncomfortable question to a person who will listen and help them decide what to do next.

Evidence behind this article

The research describes patterns of use and perceptions of trust. It does not show that every child who asks AI for advice will become dependent on it. Our practical suggestions draw on this evidence and published guidance; they are not clinical advice.

Further reading

Read more posts