Research Summary: Generative AI Purpose-built for Social and Mental Health: A Real-World Pilot

A real-world pilot study of a Generative AI foundation model, specifically designed for mental health support, demonstrated significant reductions in depressive and anxiety symptoms among 299 US adults over a 10-week period. The pilot also observed improvements in loneliness, behavioral activation, and social interaction, alongside appropriate escalation of automated safeguards by clinicians. Outcomes varied, with 42.8% of participants showing improvement or rapid improvement, linked to early working alliance and greater engagement.

Why it matters

This research provides early real-world evidence for the efficacy of purpose-built Generative AI in addressing mental health challenges, potentially expanding access to care and improving outcomes for a significant portion of the population. Understanding the factors influencing user engagement and varying response trajectories is critical for strategic deployment and refinement of AI-driven health interventions.

Key insights

• Generative AI chatbots can extend access to mental health care, as evidenced by a naturalistic pilot.
• Participants experienced a significant decrease in depression (Cohen's d 0.93) and anxiety (Cohen's d 0.79) symptoms within 10 weeks of using the AI model.
• Improvements were also observed in related areas, including loneliness, behavioral activation, and social interaction.
• Clinicians confirmed the appropriate function and escalation of automated safeguards embedded within the AI system.
• Participant outcomes varied, with 57.2% showing no response, 37.1% improving, and 5.7% rapidly improving.
• Early establishment of a working alliance with the AI and higher engagement levels were correlated with better outcomes.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2511.11689