For the first time, machine learning methods are applied to the gas diffusion layer (GDL) to characterize structural and transport properties. In the first investigation, a benchmark study of seven machine learning methods used to predict the permeability and diffusivity of GDLs in two phase flow conditions was provided. The best performing models could predict single- and two-phase permeability and diffusivity with R² >​ 0.95. In the second investigation, numerically generated image data was used to train three CNNs which predicted the porosity of X-ray images from commercial GDL materials (MAE