Digital Psychiatry with Physical Artificial Intelligence: Recent Advances and Limitations

Digital psychiatry has evolved from screen-based and virtual systems toward physically embodied artificial intelligence (AI) capable of perceiving and acting in the real world. Physical AI integrates sensors, cognitive algorithms, and actuators into closed perception-cognition-action loops, enabling dynamic and multimodal interaction with patients. This narrative review synthesizes recent advances in the technological foundations, clinical applications, and ethical implications of physical AI within psychiatry. Diagnostic applications leverage multimodal sensing and social robotics to detect psychomotor, affective, and behavioral changes in real time, offering new biomarkers for mood and cognitive disorders. Therapeutic models employ socially assistive robots and haptic systems to enhance engagement, empathy, and behavioral activation, while integrative approaches combine telepsychiatry and virtual reality to maintain continuity of care. Preliminary studies suggest potential benefits for older adults and mobility-limited populations; however, evidence remains limited by small sample sizes and methodological heterogeneity. Ethical and humanistic considerations-such as patient autonomy, privacy, safety, and relational authenticity-remain central to responsible implementation. Future research should establish standardized outcomes, regulatory frameworks, and equitable access models. As embodied AI becomes more autonomous, psychiatry must preserve clinician oversight and the therapeutic alliance, ensuring that technology functions as an adjunct rather than a substitute for human care. Physical AI thus represents both a technological milestone and an ethical inflection point, inviting psychiatry to integrate intelligence, embodiment, and compassion in advancing human-centered mental health care.