Visual tracking aims to accurately track specific objects in continuous videos, facing challenges such as occlusion, deformation, and background interference. The Transformer model demonstrates its potential in visual tracking with its excellent feature extraction and sequence modeling capabilities. However, in the pursuit of higher performance, as the design of the Transformer model continues to deepen, particularly as the number of Transformer blocks increases, a degradation problem arises, causing the model to become more sensitive to noise and interference, thereby losing the target. To overcome this challenge, we propose the Dynamic Modulation and Key Augmentation Transformer (DMKAT), which comprises two key parts: the dynamic information modulation mechanism (DIMM) and the key information enhancement decoder (KIAD). DIMM achieves dynamic regulation of information by integrating gated linear units and dynamically controls the flow of information based on the importance of input features. The proposed DIMM ensures that only key information beneficial to visual object tracking is transmitted while suppressing noise and interference. KIAD can enhance the ability to capture and utilize key information by improving the decoder structure and introducing a mechanism for directly propagating attention scores. The additional advantage is that the decoder can directly use the encoder’s high-resolution feature maps and focus on target-related areas through the attention mechanism, thereby improving the accuracy and robustness of target tracking. This dual mechanism can effectively alleviate the degradation problem. Extensive experiments on multiple standard visual tracking datasets show that the proposed model achieves improved performance compared to state-of-the-art methods.