A new method to reconstruct the microstructure of the gas diffusion layer (GDL) in fuel cells is proosed in this work to investigating the influence of fiber’s in-plane distribution on the GDL transport performance. A 3D model of GDL is obtained by threshold segmentation of the 2D slices acquired through X-ray computed tomography (XCT) scanning. Fiber-tracking technique is used to differentiate fibers and binders, to obtain information such as the in-plane orientation probability distribution of fibers, local porosity of the fiber skeleton, and the proportion of fiber and binder components as control factors. This enables the reconstruction of a more accurate GDL fiber skeleton. A pore scale model is then reconstructed by adding binders through morphological processing. Performance simulations are conducted on a 1 000 μm×1 000 μm×200 μm computational domain to analyze the effects of different fiber orientation distributions on the GDL diffusivity, electronic and thermal conductivities. Because most of the fibers of carbon paper in the manufacturing are arranged in the direction of the paper machine (machine direction), different arrangements seriously affect the performance of GDL in the machine direction, cross-machine direction, and through-plane direction (TP direction). The results show that as the concentration of fibers in the machine direction increases, the gas transmission and thermoelectric conduction performance increase in the machine direction, and decrease in the cross-machine direction. In the TP direction, the GDL model with a consistency coefficient of 0.029 for the orientation distribution of fibers concentrated in the machine direction in this study has better performance. The study reveals that the electrical conductivity and thermal conductivity are more sensitive to the fiber orientation distribution than the gas diffusion rate.