Date created: | Last Updated:

Category: Project

Description: Widespread ad hoc use of AI chatbots for mental health support highlights the need for validated AI approaches to improve psychological well-being. We conducted a preregistered, randomized experiment with a large, diverse US sample (N=2,936) testing four chatbots that implemented multi-step strategies grounded in psychological research. The chatbots guided participants to savor positive experiences, express gratitude to a close other, reflect on sources of meaning, or recast their life story as a hero’s journey. Relative to a control chatbot, all four improved five preregistered outcomes: affect, meaning in life, life satisfaction, and reductions in anxiety and depressed mood. Effects held for participants with high baseline anxiety or depressed mood. Chatbot use also increased willingness to see a human therapist, including among those previously unwilling. A nationally representative survey (N=3,056) found that half of US adults expressed interest in using empirically validated well-being chatbots. Together, these findings indicate that structured AI dialogues can improve psychological well-being at scale. During a period of rapidly expanding, yet unvalidated, use of AI for mental health support, these results point to the potential value of structured AI dialogues based on prior research as a practical and scalable path for evidence-based support.