Depression and anxiety are crucial measures for many clinical studies, both as primary and secondary outcomes. However, collecting frequent and precise mental health data important for many clinical trials can be error-prone, inconvenient, and time-consuming for the patient and may result in dropouts. Providing this data can be particularly challenging for non-native English speakers and other disadvantaged populations with poor health literacy. We have developed highly accurate AI algorithms that leverage off-the-shelf wearables to continuously measure anxiety, depression, burnout, fatigue, and sleep, correlating strongly with the gold standards for these measurements. The wearable data can also trigger an AI-based app dialogue around symptomatology, just-in-time therapy, or additional data collection. In a trial with college students and another with the general population, our algorithms accurately tracked mental symptoms, showing an NRMSE of < 0.08 for both the PHQ-8 and GAD-7 measures and an NRMSE of < 0.09 for self-reported, nine-point measurements of sleep and fatigue. Additional studies with Stanford Hospital on nurse burnout and with the US military to monitor active-duty personnel’s physical and psychological health are in progress. We are preparing a pre-submission letter to the FDA for a de-novo Class II screening device and anticipate approval in 2024.