FamilyAtlasStart by asking what the tool provides: entertainment, conversation, roleplay or a place to discuss difficult feelings. Do not assume emotional attachment is a diagnosis. AI companions can produce persuasive, intimate or inappropriate responses, and children may disclose private information to them. A friendly tone does not make the system a person, a confidential therapist or a dependable source of crisis help.
UNICEF’s current guidance identifies child-rights risks in relational AI; long-term evidence about children’s use is still developing. Review the specific provider’s age rules, privacy policy and safety settings rather than assuming every chatbot behaves alike. Set boundaries around personal details, sexual content, money and nighttime use. Keep trusted people available for real support. If the conversations involve coercion, self-harm or sexual exploitation, respond as a safety concern rather than treating it only as too much screen time.
UNICEF: When AI becomes a friend: child rights risks — 2026 policy brief · UNICEF: Policy guidance on AI for children · 988 Suicide & Crisis Lifeline: Help someone else
The parent concern
The starting question is what the child is using the companion for and what the provider allows. Emotional attachment by itself is not a diagnosis; discuss use before inferring its effect.
What UNICEF warns about—and what is not established
UNICEF’s 2026 policy brief identifies risks involving relational design, privacy and inappropriate advice or content. It is a policy synthesis, not a clinical trial showing that every child—or every companion—has the same outcome.
Long-term developmental effects are not settled. Avoid claims that a particular amount of chatbot use causes a diagnosis or that an AI friendship is always harmless.
UNICEF: When AI becomes a friend: child rights risks — 2026 policy brief
Review the actual service
Check the provider’s current minimum age, account type, data-use policy, deletion process and available safety settings. Do not give a younger child an adult account to skip a restriction.
Ask what the child has been encouraged to share or buy. Set a clear rule against names, addresses, school details, intimate images and login information. Explain that a generated response can sound certain while being wrong.
Keep human support easy to reach
Ask whether use is displacing sleep, school, friends or activities the child values. Review changes together without shaming the attachment. A child may be using the tool because a human conversation feels difficult; make the next human conversation easier.
An AI companion is not a crisis service. If the child feels unsafe, seek qualified support or emergency help rather than asking the bot to manage the risk.
What you can do right now
- Ask what the child likes about the companion.
- Check current provider age rules and privacy terms.
- Review sexual-content and purchase settings where offered.
- Agree on information never to share.
- Set a nighttime boundary and review use together.
- Name trusted adults and real support services.
A way to start the conversation
“What does talking to it help with? I am not going to mock it. I do want us to understand what it stores and make sure you have real people to help when something is serious.”
A suggested script to adapt, not a clinical intervention.
What technology cannot solve
Blocking a supported app or domain changes access. It does not evaluate every chatbot reply, detect emotional dependence or provide mental-health care.
Questions parents ask
Are AI companions proven to harm every teen?
No. Evidence is developing, and product designs and individual contexts differ. Avoid universal clinical claims.
Can the bot keep a secret like a therapist?
Do not assume confidentiality. Read the provider’s data policy and avoid disclosing sensitive information.
Should I abruptly remove it?
Address immediate danger promptly. Otherwise, understand the use and agree on specific boundaries and human support.
Sources and review
Provider documentation explains settings and eligibility. Professional guidance provides context, while our suggested family strategies are labeled separately. FamilyAtlas claims were checked against implementation, with release limits stated above.
- UNICEF — When AI becomes a friend: child rights risks — 2026 policy brief guidance · checked 2026-09-30
- UNICEF — Policy guidance on AI for children guidance · checked 2026-09-30
- 988 Suicide & Crisis Lifeline — Help someone else safety · checked 2026-09-30
Next source review: 2026-12-30. Menus, eligibility and rollouts can change; follow the linked provider instructions if your screen differs.
Educational guidance, not medical advice. The cited professional guidance is separate from our suggested family strategies. This article has not had independent clinical review; seek qualified help for persistent concerns or urgent safety risks.
Written by FamilyAtlas; product descriptions are not independent product recommendations. No claim of medical or psychological review is made.