When AI Venting Helps or Harms: How Perceived Anonymity Shapes Supportive Interactions With Conversational Agents

Artificial intelligence conversational agents are increasingly used for emotional venting, yet their psychological implications remain paradoxically unclear. AI-assisted venting can both support and harm well-being, raising questions about when human-AI interactions are beneficial and how specific design features are associated with more supportive or more stressful outcomes. Drawing on a cross-sectional survey of 408 AI chatbot users in China, this study tests a dual-pathway model. More frequent AI-assisted venting was positively associated with perceived stress directly but negatively associated with stress indirectly through perceived AI support and emotion regulation. Perceived anonymity moderated these associations: at higher levels, the positive direct venting–stress association was attenuated, whereas the negative indirect association was stronger. These findings position perceived anonymity as a design-critical boundary condition in AI-mediated interactions. The study advances HCI theory by showing how a design perception relates to psychological pathways, offering implications for designing safer, more supportive conversational agents.