Artificial intelligence is increasingly used in mental health assessment to support early detection of psychological disorders through clinical, behavioral, linguistic, physiological, and digital activity data. This article evaluates AI-assisted screening for depression, anxiety, stress-related symptoms, bipolar-risk indicators, and suicide-risk warning signs. The methodology used questionnaire scores, smartphone activity, speech-text markers, wearable-derived physiological signals, and multimodal ensemble learning. Results of the study indicate that the multimodal ensemble model performed the best, yielding an accuracy rate of 91.2%. Further, this model yielded 90.4% sensitivity, 91.8% specificity, 90.7% F1-score, and an AUC of 0.95. Disorder-wise sensitivity analysis revealed higher sensitivities for depression, anxiety, and suicide warning signs, while sensitivity for bipolar disorder was lower, likely due to the longer observation period required. Explainability analysis suggested that the presence of the following predictors was significant: irregular sleep, negative emotional content, heightened symptom severity, reduced social interaction, late night phone use, and diminished heart rate variability. These findings indicate that AI can be used as a supervised clinical decision support system.