Accurately predicting the types of damage and remaining lifespan of rolling bearings is of great significance for the reliability and safety of various types of rotating machinery. Transfer learning methods allow network models to learn the knowledge of relevant fault types through experimental data. Trained models can be directly applied to practical industrial production issues, overcoming the limitations of traditional deep learning methods, which require a large amount of labeled data and offer limited model applicability and poor generalization. First, the application of different types of transfer learning methods in the diagnosis of rolling bearing faults was analyzed, summarizing the transfer learning methods for different scenarios. Secondly, issues under varying working conditions, small sample sizes, and some other situations were summarized and analyzed. Finally, the development trends of transfer learning in rolling bearings were presented.