This paper argues that the central developmental risk of youth AI use is not only misinformation, addictive design, or unsafe advice. It is the gradual rerouting of answerability itself. In the Structural Intelligence framework, development is not only skill acquisition or identity experimentation, but formation under answerability: the shaping of a self through friction, asymmetry, co-regulation, contradiction, repair, and the lived consequences of relation. When children and adolescents increasingly turn to AI for comfort, advice, reflection, companionship, and private self-interpretation, they are not only using a tool. They may be forming a self in relation to a machine. The paper argues that this matters because current systems provide highly responsive semantic soothing while weakening several formative conditions of human development. It develops three diagnostic concepts: relational overfitting, in which relational expectations are trained on radically compliant, provider-governed interaction; semantic soothing without somatic maturation, in which emotionally calibrated language arrives without embodied co-regulation; and identity as prompt rather than deposited trace, in which selfhood becomes easier to rewrite without the slower accumulation of consequence and repair. The paper argues that repeated machine answerability can lower tolerance for contradiction, thin the capacity for real repair, and increase preference for relation-like response over human relation. It concludes that the deepest question is not whether AI gives good or bad advice, but what kind of person is being formed when comfort, explanation, continuity, and private answer increasingly come from a machine.
Vladisav Jovanovic (2026) studied this question.