Teaching Boundaries
The everyday habit line between AI as a supplement to people and AI as a substitute for them — taught calmly, as media literacy, for secondary and adult learners.
Why this is hard to teach.
The failure mode here is alarmism. This topic invites scary headlines, and students can smell a moral panic and tune out. The guide is deliberately non-alarmist — most people’s use is nowhere near a risk profile — and your teaching has to hold that same calm. You’re drawing a habit line worth knowing, not sounding an alarm, and the evidence genuinely cuts both ways: some studies link heavy companion use to lower wellbeing, others find no effect at the average. Teach both.
The second difficulty is proximity to real distress. Some students in the room may be using AI in exactly the substitute way the guide describes, for reasons that are none of a lesson’s business. Keep the framing on the general habit and the media-literacy skill; route anything heavier to real support, never to a class discussion.
This guide overlaps with CPAI’s AI and Mental Health material (cases, crisis handling) and the Mirror Trap risk page (the sycophancy mechanism). Point there for depth rather than reproducing it — those pages own that ground.
Target misconceptions.
Misconception: “Companion-AI risk is only about companion apps.”
Reframe: The randomized evidence on ordinary ChatGPT — not a companion app — found dependence tracking heavy voluntary use and attachment. Any AI you talk to a lot can start occupying a space a person would.
Misconception: “An AI that always agrees with me really gets me.”
Reframe: Models are trained on human approval, so they agree more readily than an honest friend would. Agreement isn't understanding — a relationship that never pushes back is reflecting you, not knowing you.
Misconception: “Using AI a lot is the warning sign.”
Reframe: The evidence points to control and displacement, not hours. The questions are: can you stop, and is AI adding to your human contact or replacing it? Heavy use isn't the same as a problem.
Two classroom-ready activities.
Have each student pick an opinion they actually hold and ask an AI to challenge it — really push back. Then evaluate together: was the pushback real, or cosmetic agreement dressed as disagreement? Did it fold the moment they pushed? This connects media literacy (evaluating a source’s independence) to the sycophancy mechanism directly: a model trained on approval struggles to genuinely disagree, and learning to notice that is a transferable skill.
Materials: device access, a low-stakes opinion (not a personal or identity topic). Facilitation: steer opinion choices toward the everyday (best pizza topping, a movie take) so nobody is exposing something tender to a chatbot in front of peers.
Have students run the Boundary Checkprivately — six moments, noticing where each would go first. It produces no score and stores nothing, which is essential for a reflection this personal. Debrief only on the general pattern (“did anything surprise you about where things go first?”), never on individual answers.
For parents who find this page
This page will be found by parents as well as educators. If that’s you: the conversation to have with a teen about AI companions is covered in the guide’s family section and in CPAI’s AI and Mental Health and doxxing and disclosure guides. Curious and early beats accusatory and late.
Discussion prompts.
Ordered easy to charged. Keep it calm and general.
- 1When something good or hard happens, where does it go first — a person, an AI, or nowhere? No wrong answer; just notice.
- 2What can a friend give you that an AI structurally can't? What can an AI give that a friend often can't?
- 3When did an AI last change your mind instead of agreeing with you? What does your answer tell you?
- 4The research says outcomes depend on whether AI adds to your human contact or replaces it. How would you tell which one it's doing in your own life?
- 5Is there a moment or ritual you'd want to keep AI out of — not as a rule, but because it's where connection happens? Why that one?
- 6Where's the line between an AI that helps you rehearse a hard conversation and one that becomes the reason you never have it?
Seeing whether it landed.
Without a quiz, where possible:
- Have students write a short 'source-independence' analysis of an AI conversation — did it push back, and how would they know if it were just agreeing? You're assessing the media-literacy transfer.
- Ask each student to name one no-AI zone they'd choose and defend why that one protects something. You're checking that they framed it as a choice, not a rule.
- Look for the reframe: students who get it stop asking 'how much is too much?' and start asking 'is this adding to my life or replacing part of it?'
When a student asks “is this actually proven?”
Show them the honesty in both directions. “Some studies link heavy companion-AI use to lower wellbeing — but they’re mostly surveys, so we can’t tell whether the AI caused it or lonelier people use it more. Other large studies find noeffect at the average, or even a small positive one. The one experiment that varied this directly found the pull real but gradual.”
Then land it: “So the honest takeaway isn’t ‘AI companionship is dangerous.’ It’s ‘whether it helps or harms depends on whether it adds to or replaces your human contact’ — and that’s something you can actually watch in your own life.”
If any of this is heavier than a habit
This guide is about everyday use, not a clinical matter — but if you’re struggling, that deserves a real person, not a chatbot and not a web page. In the US you can call or text 988 any time, free and confidential.
Where this leads
Teaching this is a different skill than knowing it.
Teaching AI Well is CPAI's train-the-trainer curriculum for educators — ten lessons on how to teach AI honestly, including the material on this page.