Abstract: Ultrasonic guided wave inspection of rail defects is a research focus in the field of railway track maintenance, but the presence of temperature stress affects the dispersion characteristics of guided waves, thus impacting the effectiveness of detection, obtaining so the dispersion characteristics of guided waves under temperature stress and selecting appropriate detection modes can contribute to improving detection accuracy. Therefore, this paper establishes a model for the propagation of guided waves in rails under temperature stress, solves for the dispersion characteristics of guided waves, and analyzes the relationship between temperature stress and group velocity. Based on this, criteria for selecting detection modes are investigated to choose guided wave modes suitable for rail inspection under temperature stress. Firstly, temperature stress is equivalently considered as axial stress, and a model for the propagation of ultrasonic guided waves under temperature stress is established using the semi-analytical finite element method (SAFEM). The dispersion characteristics of guided waves in rails under temperature stress are obtained. Secondly, a method for selecting detection modes is constructed based on frequency, rail bottom vibration amplitude, mode significance factor, and stress sensitivity factor to determine the optimal detection mode. Finally, verification and analysis are conducted through finite element simulation and on-site experiments. The results indicate a good linear relationship between temperature stress and rail guided wave group velocity. As temperature stress changes, the group velocities of different modes vary, albeit with different magnitudes. The optimal detection mode in the rail waist region is Mode 3, which exhibits strong discernibility and is less affected by temperature stress. These findings enhance the accuracy of ultrasonic guided wave detection of seamless rail defects under temperature stress and hold practical significance.