Real-World Engagement With a Generative AI Conversational Agent for Mental Health Support: Retrospective Descriptive...

Area of Science:

• Digital mental health interventions
• Artificial intelligence in healthcare
• User engagement studies

Background:

• Generative artificial intelligence (GenAI) conversational agents are increasingly integrated into digital mental health interventions (DMHIs).
• Empirical data on real-world engagement, usage patterns, and satisfaction with these agents are limited.

Purpose of the Study:

• To characterize users of the GenAI conversational agent within the Mental DMHI.
• To examine real-world usage patterns, satisfaction, and feedback.
• To explore predictors of engagement with the GenAI conversational agent.

Main Methods:

• Retrospective analysis of naturalistic user data from 5082 paid subscribers (October 2024 - March 2026).
• Collected onboarding characteristics and session satisfaction via app-native items.
• Captured session-level engagement metrics through backend data; analyzed using descriptive statistics, ANOVAs, t-tests, and mixed-effects logistic regression.

Main Results:

• High user engagement with 59,602 sessions, primarily in evenings and outside business hours.
• High mean session satisfaction (4.5/5), with "Insightful," "Felt seen," and "Good advice" as top descriptors.
• High session-to-session return rate (92.6%), with satisfaction predicting return (OR 1.35).

Conclusions:

• GenAI conversational agents in DMHIs demonstrate sustained real-world engagement, particularly outside traditional care hours.
• Objective engagement data suggest GenAI may support individuals facing barriers to traditional mental health services.
• Future research should investigate if these engagement patterns lead to clinically meaningful outcomes.