Early intervention in children with developmental delay is crucial to optimize long-term prognosis, but traditional rehabilitation has long been trapped by uneven distribution of professional resources. Information and communication technology enables remote intervention in home scenarios, while the deep embedding of artificial intelligence (AI) is fundamentally transforming the field paradigm. This paper focuses on the application of artificial intelligence-enabled telerehabilitation in home intervention of children with developmental delay, and provides a panoramic review from technical base to clinical landing. This paper explains how the core technologies such as computer vision, machine learning and multi-modal fusion support early recognition, quantitative evaluation, individualized scheme generation and dynamic tracking, and then deconstructs the platform architecture logic, algorithm iteration path and data security from the perspective of system engineering. Moreover, through the comparison of representative clinical prototypes and commercial platforms, the application scenarios and functional boundaries of each technical route are refined. However, despite it possesses tremendous potential, this field still has outstanding shortcomings in the dimensions of clinical evidence-based verification, model generalization, privacy ethics and payment mechanism. Therefore, the key to future development lies in building a closed loop of “clinical demand traction-technological innovation drive-industrial value transformation” to form evidence-based practice standards and inclusive service system.