This systematic review and meta-analysis investigated the performance of wearable artificial intelligence (AI) in detecting and predicting depression. 54 studies were included, revealing a pooled mean highest accuracy of 0.89, sensitivity of 0.87, specificity of 0.93, and RMSE of 4.55. Subgroup analyses showed statistically significant differences in performance across algorithms and wearable devices. While promising, wearable AI is not yet ready for clinical practice and requires further research to improve its performance and distinguish depression from other conditions.
