Generalized Anxiety Disorder (GAD) is a common psychiatric disorder, affecting approximately 3–6% of the population worldwide and about 0.57% in India, over the last few decades it has gradually increased. Conventional interventions, such as pharmacological treatments, psychotherapy, and self-reporting mobile apps, may not always ensure prompt, real-time support during acute anxiety episodes. Moreover, so far, all digital solutions have relied on single-sensor approaches, such as galvanic skin response or heart rate, which can give incorrect or incomplete assessments due to external interferences and the complex, multi-dimensional nature of anxiety. In this paper, we propose a novel AI-driven multi-sensor fusion system incorporating facial recognition, galvanic skin response, temperature, and accelerometer signals for continuous monitoring of emotional and physiological states. Unlike these unimodal systems, the proposed framework offers real-time robust detection of distress while providing personalized, non-invasive therapeutic interventions through prompts for relaxation, biofeedback, and guided breathing exercises. Significantly, the design ensures uninterrupted support by operating independently of facial recognition in low-light or privacy-sensitive environments. By integrating adaptive learning algorithms across a diverse range of data sources, the approach enhances accuracy, reduces algorithmic bias, and, over time, adjusts responses to individual user profiles. This constitutes one of the significant advances in digital mental health care, as it switches from reactive treatment to proactive treatment, context-aware management of GAD, reducing the risk of escalation and improving the quality of life, thereby contributing to the global effort toward mental well-being.