Purpose <p>Artificial intelligence (AI) has the potential to transform psychiatric care, offering a solution to complex problems such as predicting illness prognosis, facilitating diagnostic decisions, and personalizing treatment plans. Despite growing enthusiasm surrounding these innovations, widespread integration of AI technologies into clinical practice remains a challenge. The purpose of this review is to highlight opportunities, applications and challenges of AI in psychiatry, by focusing on recent research across three domains: (1) smartphone sensing and digital phenotyping; (2) chatbots; and (3) natural language processing.</p> Recent Findings <p>Accumulating evidence shows that multimodal data streams captured through smartphone sensors, electronic heath records, and textual information derived from social media posts and session transcripts can be leveraged to build machine learning models capable of predicting patient outcomes. The efficacy of generative AI chatbots on psychiatric symptoms is also emerging, though reporting guidelines on harms and safety standards are urgently needed.</p> Summary <p>The preponderance of pilot/feasibility studies in this field suggests that research should now shift focus towards validation to help determine clinical relevance and inform patient decisions in real-world practice. Doing so could expedite the safe integration of AI in psychiatric care and ensure that each patient receives appropriate services personalized to their needs.</p>