What is the short answer?
An AI companion is built around conversation that feels personal. Before allowing its use, review the app's purpose, age requirements and how a young person would raise concerns with an adult. A warm response does not make a chatbot a trusted person.
What does the 2025 study actually tell us?
Common Sense Media surveyed 1,060 US teenagers aged 13–17 in April and May 2025. Seventy-two percent reported using AI companions at least once. The report concerns conversational companionship, not every educational use of AI.
This is a self-report usage survey. It does not establish that 72% experienced harm, or justify applying that percentage to elementary-age children or every country. Its date matters too: it describes 2025.
Source: Robb & Mann / Common Sense Media — Talk, Trust, and Trade-Offs (2025), pp. 1–2
How can a family discuss this without an interrogation?
Ask about the difference between a tool that helps with a task and one that encourages an ongoing conversation. Fictional examples let families discuss the distinction without demanding to see a teenager's private conversation.
Try three questions: what appeals to you about it? What would you do if it asked you to keep a secret? Who would you turn to if a response worried you? The purpose is to establish a clear route to help, not to obtain the answer an adult wants to hear.
What boundaries can we agree on at home?
An agreement should be easy to explain through examples. 'If a conversation makes me uncomfortable, I close it and tell you' is easier to practice than 'be responsible online.' Revisit the agreement when the tool or the circumstances change.
- Review an app's purpose and requirements together before use.
- Keep private information about yourself and other people out.
- Agree on times for use and a simple way to end a conversation.
- Name a real person to contact when something causes concern.
What can a school practice without opening accounts?
Offer a fictional dialogue on paper: a character receives a friendly answer that asks for personal information. The group identifies what invites trust, what information should stay private and which adult could help.
Label the dialogue as fictional. There is no need to display students' private conversations or ask for personal disclosures. Focus on observable decisions: pause, identify the request and choose someone to consult.
How could an AI with Marcé session address this?
Defend Yourself Against AI gives groups a way to practice privacy, judgment and asking for help using age-adapted examples. A school can raise this topic in the initial conversation so that the scope fits its needs. The session is not therapy or a replacement for support from families and school staff.
Defend Yourself Against AI. Explore the program and its PDF →Which sources did we consult?
How do we put these answers together?
We check factual claims against the linked sources. Activities are educational suggestions from AI with Marcé, not research findings or undocumented classroom experiences. Dates indicate publication and content changes. Email us to suggest a correction. Email Alex.