Abstract: Based on a review of different nondestructive testing methods, we focused on three ultrasonic testing methods for composite insulator defects and analyzed their respective advantages and disadvantages. We also discussed the ultrasonic defect identification method based on the machine learning principle, summarized the advantages of deep learning in ultrasonic defect recognition and prediction, and discussed the three major problems in the ultrasonic testing of composite insulators based on deep learning. Finally, a feasible solution strategy is given, which can provide a technical reference for further research in this field.