The rapid evolution of smart cities demands highly efficient, intelligent, and adaptive sensing networks to optimize urban infrastructure, environmental monitoring, and resource management. Conventional IoT sensor networks often suffer from limitations such as high-power consumption, restricted sensitivity, and inefficient data processing. This research presents a metasurface-enabled AIoT (Artificial Intelligence of Things) sensor framework, integrating reconfigurable metasurfaces with AI-driven analytics to enhance urban sensing, communication, and automation. Metasurfaces, engineered nanostructures capable of manipulating electromagnetic waves, enable ultra-sensitive, programmable, and energy-efficient sensors. These sensors adapt to environmental changes dynamically, offering superior real-time data acquisition for applications such as air quality monitoring, structural health assessment, intelligent traffic management, and smart grid optimization. Integrating AI-based edge computing enhances real-time data processing, reducing latency and computational overhead, ensuring seamless operation even in high-density urban environments. Furthermore, the proposed architecture leverages deep learning models and predictive analytics to facilitate anomaly detection, early warning systems, and autonomous decision-making for urban sustainability. A novel hybrid computational framework combining metamaterial physics with deep neural networks is developed, demonstrating superior sensing accuracy, reconfigurability, and resilience compared to conventional IoT sensors. Experimental simulations and prototype validations confirm the system’s scalability, robustness, and real-world feasibility. By addressing key challenges in sensor adaptability, efficiency, and AI-driven automation, this research establishes a transformative approach to smart city development, ensuring sustainable, resilient, and intelligent urban ecosystems.