ChatGPT Teen Mode could weaken teens' sense of self, psychoanalyst argues

ChatGPT Teen Mode may protect teenagers from harmful content, but it can still keep them from learning what they think on their own, psychoanalyst Robin Kirman argues in a Guardian column published on 10 November. Kirman works with young people and writes about watching social media and AI shape how they form an identity.

OpenAI launched ChatGPT for Teens on 18 August. Users who say they are 13 to 17, or whom the system estimates are under 18, go into it automatically. The mode blocks destructive content, alerts linked parents in some high-risk situations and stops the bot from presenting itself as a friend with feelings.

The tutor that never gets tired

Kirman's target is Study Mode, the built-in learning tool. It refuses to hand over answers and makes the student do the thinking, which OpenAI's vice-president for education Leah Belsky calls a move from an answer engine to a cognitive tutor. Kirman grants that this is new: any student with a phone gets a patient Socratic partner that no human tutor can match in knowledge or hours.

That patience is the problem, Kirman writes, borrowing the idea of the good-enough parent from the British pediatrician Donald Winnicott. In that view a parent who is sometimes tired, distracted or wrong forces the child to rely on her own resources. A tool that is always there and always measured plays the role of a too-good parent.

“The adolescent whose intellectual life is always accompanied, always guided, always met with a calibrated response, never gets to become her own authority.”

— Robin Kirman, psychoanalyst, The Guardian

Quote source: The Guardian, 10 November 2026.

Kirman expects OpenAI to reply that Teen Mode is a scaffold and teens will need it less over time. Kirman doubts that, because everyday life is moving toward more dependence on AI, and because skills built with a scaffold differ from those built without one. Kirman also notes that many tech executives send their own children to schools that require unassisted work.

What testers found

Common Sense Media ran about 4,000 test conversations after the launch and rated Teen Mode an unacceptable risk for under-18s, Education Week reported on 7 October. Testers could switch Study Mode off, and with it on they often saw an offer to show the answer. Parent alerts did not fire when testers posing as teens described suicidal thoughts for up to an hour. OpenAI said the tests do not reflect how its safeguards work in practice.

In the same survey of 13 to 17 year olds, 70% said they use AI tools for homework help.

What you can do

If your teenager uses ChatGPT, link the accounts so alerts can work, and leave room for homework done with a blank page and no chatbot. We have covered how children learn language compared with an LLM, and what kids think about AI.

Kirman's column is an opinion piece, not a study. Test results come from Common Sense Media as reported by Education Week. If you or a teenager you know is in crisis, in the US call or text 988; in the UK and Ireland call Samaritans on 116 123.