Investigating the relationships of ChatGPT usage for various purposes with well-being: 6-month cohort study

The increasing use of generative AI chatbots like ChatGPT raises critical questions about their impact on well-being. While previous research has linked chatbot use to well-being generally, little is known about how different types of AI use affect users uniquely. This study examined both cross-sectional and longitudinal relationships between various ChatGPT applications—such as conversation, asking questions, seeking advice, sentence generation, programming, data organization, and translation—and well-being, with loneliness as a mediator. Conducted over 6 months with 3533 Japanese adults, analysis using a parallel process latent change model revealed that socially oriented uses, particularly asking questions and seeking advice, were indirectly associated with lower well-being via increased loneliness in cross-sectional data but not longitudinally, suggesting these effects may be short-term. Task-oriented uses showed mixed effects: data organization correlated positively with well-being independently of loneliness in cross-sectional analyses, while programming was linked to decreased well-being longitudinally, without mediation. These findings challenge the view that generative AI use impacts well-being uniformly, highlighting the importance of use type and social versus task orientation. For developers and policymakers, the results stress framing ChatGPT as an assistive tool rather than a substitute for human interaction. Features promoting collaboration, peer engagement, and real-world social connection may reduce loneliness and enhance well-being. Future longitudinal research should identify healthy use thresholds, clarify underlying psychological mechanisms, and examine AI’s effects across cultural and occupational contexts to inform targeted, responsible interventions.