A pore network model (PNM) is proposed for the simulation of water transport inside the cathode side gas diffusion layer (GDL) of polymer electrolyte fuel cells (PEFCs) during the transient start-up period as well as the steady state. Numerous two-dimensional random networks representing GDL are generated followed by statistical averaging of the results (Monte Carlo methods) to circumvent the uncertainties imposed by random pore size distributions. The resulting liquid water saturation profiles within GDLs exhibit concave patterns which is typically encountered in capillary fingering flow regimes in porous media. The effect of GDL thickness and current collector rib width as two geometric parameters on water transport dynamics are separately investigated. It turns out that thin and thick GDLs compared to the base case can have contradicting outcomes on the account of total water saturation in the network. On the other hand, wide current collector ribs give rise to liquid water saturation and build-up within GDL which can lead to flooding. At the end, three-dimensional networks are generated demonstrating higher pore connectivity which results in higher percolation times and different invasion patterns.