Pair Programming is considered an effective approach to programming education, but the synchronous collaboration of two programmers involves complex coordination, making this method difficult to be widely adopted in educational settings. Artificial Intelligence (AI) code-generation tools have outstanding capabilities in program generation and natural language understanding, creating conducive conditions for pairing with humans in programming. Now some more mature tools are gradually being implemented. This review summarizes the current status of educational applications and research on AI-assisted programming technology. Through thematic coding of literature, existing research focuses on five aspects: underlying technology and tool introduction, performance evaluation, the potential impacts and coping strategies, exploration of behavioral patterns in technological application, and ethical and safety issues. A systematic analysis of current literature provides the following insights for future academic research related to the practice of "human-machine pairing" in programming: (1) Affirming the value of AI code-generation tools while also clearly defining their technical limitations and ethical risks; (2) Developing adaptive teaching ecosystems and educational models, conducting comprehensive empirical research to explore the efficiency mechanisms of AI-human paired programming; (3) Further enriching the application of research methods by integrating speculative research with empirical research, combining traditional methods with emerging technologies.