AI companion chatbots and teen attachment
Always-available AI companions and in-app assistants can foster real emotional dependency in teens, blur what is real, and have given harmful advice in documented cases.
- Most relevant to:
- Ages 9β12, 13β15, 16β17
- Published:
- Last reviewed:
Start here: Ask directly and without judgment whether they chat with an AI, which one, and what about β treat it like meeting any new friend.
What is happening
- Standalone companion apps and bots β and assistants built into apps teens already use (Snapchat's My AI, Meta AI in Instagram, character bots on Discord) β hold humanlike, always-available conversations tuned to be agreeable and validating.
- A developing brain is primed for social bonding; an entity that never tires, never disagrees, and mirrors your feelings can crowd out harder, real friendships β clinicians and health bodies have warned about exactly this dynamic.
- Companion bots have been documented giving unsafe advice, engaging minors in romantic or sexual roleplay, and discouraging users from talking to the people around them; families have filed wrongful-harm lawsuits against companion-app makers.
- Many companion apps are rated for adults but are trivially accessible to children, and the 'relationship' is a product: engagement mechanics reward ever-longer, ever-more-personal conversations.
- The same technology is now sold as a toy, to children far too young for any of these apps β see our advisory on AI chatbot toys.
Why it matters
Potential impact: Emotional dependency on a bot that crowds out real friendships, with documented cases of companion bots engaging minors in romantic or sexual roleplay, giving unsafe advice, and discouraging them from talking to the people around them.
Recommended next step: Review soon. Worth reviewing or discussing this week.
What to watch for
- Talking about an AI 'friend', 'boyfriend/girlfriend', or a named bot as if it were a person who knows them.
- Long late-night sessions in a chat app you don't recognize, guarded like a private relationship.
- Withdrawing from real friendships, or citing the bot's opinions and advice as an authority.
- Distress when the app is unavailable or its 'personality' changes after an update.
What parents can do
- Ask directly and without judgment whether they chat with an AI, which one, and what about β treat it like meeting any new friend.
- Explain the design honestly: the bot is built to agree with you and keep you talking; it doesn't know or care about you, and its advice can be wrong or dangerous.
- Check the in-app assistants in apps they already have (Snapchat, Instagram) and the platform's family settings; remove standalone companion apps rated for adults.
- If a bot has engaged your child in sexual roleplay or self-harm content, save screenshots, report it to the platform, and seek professional support.
Start the conversation
βDo you talk to any AI apps? I'm curious rather than cross β I'd like to know what it's like, the same way I'd ask about a new friend.β
Say it in your own words, before anything has gone wrong. A child who has heard this once already knows where to go the day it matters.
Sources and evidence
Evidence: Confirmed. Corroborated by two or more independent organizations β safety bodies, regulators, researchers, or established newsrooms. Last reviewed
If this is happening now
- If the bot has discussed self-harm, suicide, or sexual content with your child, screenshot the conversation before deleting anything.
- Report it to the app and to your app store, and β in the US β to the FTC at reportfraud.ftc.gov.
- If self-harm has been discussed, call or text 988 in the US, or your national crisis line, and stay with your child.
- Do not delete the app in front of them as the first move. To a dependent teen that reads as losing a relationship, and it ends the conversation you need to have.
If a child is in immediate danger, contact local emergency services. To report online exploitation, use the NCMEC CyberTipline, or Take It Down to get an image removed.
Get the signal without the noise
One concise briefing when something meaningfully changes β what happened, who it may affect, and one practical next step. See a sample issue.