This review explores the integration of artificial intelligence (AI) with metasurface research, highlighting the potential of AI to enhance the design, optimization, and application of metasurfaces. Metasurfaces, which are planar optical elements capable of controlling light on the sub-wavelength scale, have shown great promise in various optical applications due to their ability to manipulate electromagnetic waves with high precision. However, traditional metasurface design methods face challenges such as limited computational resources, complex optimization problems, and the need for extensive manual intervention. AI, particularly machine learning (ML), offers a powerful solution by enabling data-driven design, accelerating the optimization process, and improving the accuracy of metasurface performance predictions. The review discusses various AI techniques, including supervised and unsupervised learning, and their applications in metasurface design. Supervised learning methods such as Support Vector Machines (SVM), k-Nearest Neighbor (kNN), and Random Forest (RF) are used to predict metasurface properties, optimize structural parameters, and enhance performance. For example, SVM is employed to design reconfigurable metasurfaces based on diffraction patterns, while kNN is used to optimize the structures of metasurfaces for radiative cooling and electromagnetic absorption. RF is shown to be effective in predicting the absorption bandwidth of patterned graphene metasurfaces, achieving high accuracy and robustness. Unsupervised learning techniques are also discussed, focusing on their ability to discover hidden patterns and structures in metasurface data without the need for labeled datasets. These methods are particularly useful in exploring the vast parameter space of metasurface design and identifying optimal configurations. The review also highlights the advantages of AI in metasurface research, including its ability to handle complex optimization problems, reduce computational costs, and improve the efficiency of design processes. By leveraging AI, researchers can overcome the limitations of traditional methods and push the boundaries of metasurface technology, enabling new applications in optical imaging, sensing, and communication systems. The integration of AI with metasurface research represents a promising direction for future advancements in micro-nano optics.This review explores the integration of artificial intelligence (AI) with metasurface research, highlighting the potential of AI to enhance the design, optimization, and application of metasurfaces. Metasurfaces, which are planar optical elements capable of controlling light on the sub-wavelength scale, have shown great promise in various optical applications due to their ability to manipulate electromagnetic waves with high precision. However, traditional metasurface design methods face challenges such as limited computational resources, complex optimization problems, and the need for extensive manual intervention. AI, particularly machine learning (ML), offers a powerful solution by enabling data-driven design, accelerating the optimization process, and improving the accuracy of metasurface performance predictions. The review discusses various AI techniques, including supervised and unsupervised learning, and their applications in metasurface design. Supervised learning methods such as Support Vector Machines (SVM), k-Nearest Neighbor (kNN), and Random Forest (RF) are used to predict metasurface properties, optimize structural parameters, and enhance performance. For example, SVM is employed to design reconfigurable metasurfaces based on diffraction patterns, while kNN is used to optimize the structures of metasurfaces for radiative cooling and electromagnetic absorption. RF is shown to be effective in predicting the absorption bandwidth of patterned graphene metasurfaces, achieving high accuracy and robustness. Unsupervised learning techniques are also discussed, focusing on their ability to discover hidden patterns and structures in metasurface data without the need for labeled datasets. These methods are particularly useful in exploring the vast parameter space of metasurface design and identifying optimal configurations. The review also highlights the advantages of AI in metasurface research, including its ability to handle complex optimization problems, reduce computational costs, and improve the efficiency of design processes. By leveraging AI, researchers can overcome the limitations of traditional methods and push the boundaries of metasurface technology, enabling new applications in optical imaging, sensing, and communication systems. The integration of AI with metasurface research represents a promising direction for future advancements in micro-nano optics.