AI Chatbot Trained for Mental Health Shows Promise in Year-Long Pilot Study

Researchers have demonstrated that a purpose-built generative AI chatbot can help reduce symptoms of depression and anxiety in real-world settings, offering initial evidence for the viability of such tools in expanding access to mental health support.

Study Design and Outcomes#

In a single-arm naturalistic pilot, investigators followed 299 US adults experiencing at least moderate depressive or anxiety symptoms for up to 12 months of interaction with a foundation model specifically trained for mental health applications. Within the first 10 weeks, participants showed statistically meaningful reductions in both depression (Cohen's d 0.93) and anxiety (Cohen's d 0.79). Beyond symptom relief, users also reported improvements in loneliness, behavioral activation, and social engagement over the study period.

Variation in Response Trajectories#

Not all participants responded equally. The cohort split into three distinct groups: over half (57.2%) showed minimal change despite using the system, while 37.1% demonstrated steady improvement, and a small subset of 5.7% experienced rapid symptom reduction. Early therapeutic connection and consistent engagement with the platform emerged as predictors of better outcomes.

Clinical Oversight and Adaptive Content#

The system deployed ten distinct intervention strategies, with delivery tailored to each user's baseline anxiety and depression levels. Notably, the platform adjusted its balance of clinical versus conversational content based on symptom severity—without requiring explicit survey data—and clinicians confirmed that automated safety escalations triggered appropriately when warranted.

Implications for Digital Mental Health#

These findings provide preliminary support for AI systems purpose-built to address mental health conditions as a mechanism for broadening care access. The results suggest that specialized training, adaptive intervention delivery, and robust safety monitoring can combine to create a functional digital therapeutic tool.