With the fusion of Generative AI—particularly large language models (LLMs), multi-modal language models (MLMs) such as vision–language models (VLMs), and large action models (LAMs)—with the Artificial Intelligence of Things (AIoT), the Generative Internet of Things (GIoT) has emerged as an evolution of traditional IoT systems. The GIoT paradigm introduces advanced cognitive intelligence, automation, and sophisticated human–machine interaction, enhancing capabilities in sectors such as Industry 4.0, healthcare, and smart cities. While offering substantial potential for innovation and efficiency, GIoT remains in its early stages, facing challenges such as trust and security, efficient edge deployment, and the lack of tailored reasoning and architectural frameworks. This paper examines both the opportunities and challenges of GIoT, presenting a comparative analysis of representative architectures and assessing their implications for key performance metrics, including inference time, accuracy, scalability, and energy efficiency. Furthermore, it proposes an enhanced reasoning methodology that extends the widely adopted Chain of Thought (CoT) framework in LLMs to better address the dynamic and heterogeneous problem spaces inherent in GIoT applications. The capabilities of GIoT and the practical benefits of these contributions are illustrated through two case studies: (i) drone-based disaster management, demonstrating real-time situational awareness and decision support, and (ii) smart home systems, highlighting improved personalization, automation, and energy optimization. Collectively, these studies underscore GIoT’s transformative potential to advance operational capabilities and enrich user experiences across diverse domains.
Firouzi et al. (Sun,) studied this question.