Leveraging IoT and LLM for Depression and Anxiety Disorders: A Privacy Preserving Perspective

ABSTRACT Depression and anxiety disorders are among the most widespread and severely impairing mental health conditions worldwide, causing heavy personal and social consequences. Even if Large Language Models (LLMs) proved to be highly effective in supporting mental health, their current applications still lack contextual awareness and real‐time adaptability. The integration of Internet of Things (IoT) systems, which enable continuous monitoring of behavioral and physiological signals, can address these limitations and open the door to more dynamic and personalized assistance. In this article, we present a comprehensive survey of existing research at the intersection of LLMs, IoT, and mental health, with a particular focus on depression and anxiety. We identify a gap in current state of the art where LLM‐IoT integration remains underexplored for mental health use cases. To address this, we propose a three‐layer architecture that supports context‐aware and privacy‐preserving support. We introduce a taxonomy of system functionalities categorized into prevention, detection, alert, and action. Furthermore, we analyze the security and privacy challenges introduced by this integration and discuss mitigation strategies based on, among others, differential privacy, blockchain, and lightweight cryptography. Our findings clearly indicate the significant potential of IoT–LLM synergy in managing depression and anxiety disorders, provided that strong security and privacy safeguards are ensured prior to large‐scale deployment.