"I am here for you": How relational conversational AI appeals to adolescents, especially those who are socially and emotionally vulnerable

Conversational style powerfully shapes how adolescents perceive AI: When chatbots used relational language (first-person, warm, commitment-focused), adolescents rated them as significantly more human-like, likable, and trustworthy compared to transparent chatbots that explicitly acknowledged their artificial nature—despite both styles being perceived as equally helpful.

Vulnerable youth are most susceptible to emotional reliance on AI: Adolescents with lower quality family and peer relationships, higher stress, and higher anxiety were notably more likely to prefer the relational chatbot style, suggesting that emotionally vulnerable young people may be at particular risk for developing unhealthy emotional dependence on conversational AI.

Parents and adolescents have conflicting preferences about AI design: While adolescents preferred relational chatbot interactions, parents were more likely to prefer the transparent style that emphasized the AI's nonhuman nature, highlighting a potential gap between youth preferences and parental safety concerns.

Relational design features create emotional closeness, not just better conversations: The study found that relational framing increased adolescents' sense of emotional closeness and trust in the chatbot, independent of whether it actually provided better advice or support—meaning design choices can manipulate emotional connection rather than improve functionality.

Conversational style is a critical lever for AI safety policy: The findings suggest that how companies design chatbot interactions—particularly the language and relational cues they use—should be considered a key safety issue for youth-facing AI, warranting design guidelines and potentially regulation.