With the continuous advancement of Internet of Things (IoT) technology, the security and privacy of devices have expanded significantly, leading to an increasingly severe security situation. Intrusion detection technology has become more critical in the field of network security. As IoT evolves towards intelligent IoT (AIoT), AI technology has once again come to the forefront and has become an important tool in intrusion detection. This paper provides a comprehensive review of AI-based intrusion detection techniques. It begins by introducing key concepts related to intrusion detection, including the classification of intrusion detection systems, commonly used datasets, and evaluation metrics required for relevant experiments. The focus then shifts to machine learning and deep learning techniques, analyzing supervised, unsupervised, and semi-supervised intrusion detection methods. Additionally, it examines research on algorithm compression techniques suitable for IoT. Finally, the paper discusses the current challenges in IoT intrusion detection and outlines future research directions.