Product design question

Design the end-to-end ChatGPT experience for users under 18. How would you segment users by age and risk level, what product changes would you make for each segment, and what tradeoffs would you make between usefulness, autonomy, parental involvement, and safety?

Practice this question out loud. An AI interviewer asks it, follows up like a real interviewer would, and scores your answer. Type or speak.

What this question tests

Product design for a high-stakes, age-segmented experience balancing usefulness against real safety risk across a wide developmental range.

Segment by age and inferred risk rather than a single under-18 bucket, since a 17-year-old and a 13-year-old have very different needs and risks.
For younger or higher-risk segments, apply stronger content restrictions, more conservative default settings, and parental visibility by default rather than opt-in.
For older teens with lower inferred risk, allow more autonomy and fewer restrictions, closer to the adult experience, while still blocking clearly harmful content categories for everyone under 18.
Design age inference to combine self-reported age with behavioral and contextual signals, since self-report alone is easy to falsify and cannot be the only gate.
Set the guiding tradeoff explicitly: default toward more restriction and more parental involvement when signals are uncertain, and loosen only as confidence in the user's age and risk level increases.

Segments by age and risk rather than treating everyone under 18 identically, which is the central design challenge the question is testing for.
Proposes a concrete mechanism for handling uncertain age signals, defaulting to more restriction until confidence increases, rather than picking one static policy for the whole group.
Names parental involvement as a tunable dial tied to age and risk rather than an all-or-nothing feature.
States the core tradeoff explicitly, usefulness and autonomy versus safety and parental involvement, and gives a clear default direction when evidence is ambiguous.

Designing one flat experience for all under-18 users, ignoring the large developmental gap between young teens and near-adults.
Relying solely on self-reported age without any behavioral or contextual corroboration.

How would you handle a user who is clearly under 18 but has lied about their age?
What would you do if age-inference signals conflicted with what the user self-reported?