Too human to resist: AI Anthropomorphism and manipulation vulnerability in Gen Z

Generation Z has grown up with artificial intelligence as a daily companion, yet anthropomorphism (attributing human characteristics to non-human entities) shapes how they perceive and interact with AI systems. While anthropomorphic perceptions can enhance engagement, they create a vulnerability: when adolescents perceive AI as having intentions, emotions, or moral agency, they become susceptible to manipulation through emotionally persuasive or deceptive AI design. This study explores how adolescents anthropomorphize AI across four dimensions: cognitive (agency, intelligence), emotional (empathy, social connection), moral (responsibility, threat perception), and technological/instrumental (tool-like attributes); with particular attention to how these perceptions relate to manipulation awareness.

Using a two-phase mixed-method design, Generation Z participants (n=125) completed free association tasks, spontaneously responding to "artificial intelligence." Analysis of these responses informed the design of our second phase: four focus groups (11-15 participants each) where adolescents engaged in discussions, brainstorming exercises, and card-sorting activities according to how well given attributes fit with AI. Qualitative coding followed established theoretical frameworks on multidimensional anthropomorphism.

All four anthropomorphism dimensions emerged across both phases. Participants frequently attributed intentionality and deceptive capabilities to AI, viewing it as potentially manipulative; meanwhile simultaneously expressing trust and emotional connection. This combination is particularly concerningas emotional anthropomorphism paired with moral agency attribution creates heightened manipulation vulnerability. Focus groups revealed varied levels of critical awareness regarding risks. Some participants expressed explicit concerns about being manipulated or developing cognitive dependency, while others described deeply personal, trusting relationships with AI—using it for emotional support and personal advice—without articulating such concerns. Despite extensive AI experience, Generation Z demonstrates persistent anthropomorphism with highly variable risk awareness: recognition of potential harms is neither universal nor consistently protective. These findings highlight that AI literacy education must address not only risk recognition but also the emotional and habitual factors that sustain anthropomorphic engagement.