Psychological vulnerability in conversational AI: how incentive and urgency manipulations alter trust formation and privacy behaviour across healthcare, financial, commercial, and security contexts

Users increasingly interact with conversational AI systems in sensitive domains, yet little is known about how psychological manipulation techniques systematically alter privacy decision-making and trust formation across different contexts. We examined how two manipulation mechanisms, incentive-based (Need & Greed) and urgency-based (Time), influence user behaviour in healthcare, financial, commercial, and home security interactions.In a within-subjects experimental study (N=30[jls-end-space/]) using an interactive, scenario-based methodology, we investigated how context shapes vulnerability to manipulation. Participants engaged with control, Need & Greed, and Time conditions across four domains whilst we measured trust, comfort levels, willingness to disclose information, and perception of privacy and security.Both manipulation techniques significantly reduced comfort (p<0.001[jls-end-space/]) and increased information disclosure (p<0.05[jls-end-space/]), but context moderated these effects substantially. Time pressure increased trust in healthcare contexts whilst decreasing it in financial domains, revealing domain-specific vulnerability patterns. Qualitative analysis identified psychological mechanisms underlying these effects: prior digital experiences, agent characteristics, urgency sensitivity, and context-dependent risk assessment.These findings demonstrate that psychological vulnerability to manipulation operates through context-specific pathways, challenging one-size-fits-all protection approaches. The research provides empirical evidence for developing domain-aware security frameworks that address psychological dimensions alongside technical protections, advancing both human-computer interaction theory and practical system design.