The regular nanoporous structure make zeolite membranes attractive candidates for separating molecules on the basis of differences in transport rates (diffusion). Since improvements in synthesis have led to membranes as thin as several hundred nanometers by now, the slow transport in the boundary layer separating bulk gas and core of the nanoporous membrane is becoming increasingly important. Therefore, we investigate the predictability of the coefficient quantifying this local process, the surface permeability α, by means of a two-scale simulation approach. Methane tracer-release from the one-dimensional nanopores of an AFI-type zeolite is employed. Besides a pitfall in determining α on the basis of tracer exchange, we, importantly, present an accurate prediction of the surface permeability using readily available information from molecular simulations. Moreover, we show that the prediction is strongly influenced by the degree of detail with which the boundary region is modeled. It turns out that not accounting for the fact that molecules aiming to escape the host structure must indeed overcome two boundary regions yields too large a permeability by a factor of 1.7–3.3, depending on the temperature. Finally, our results have far-reaching implications for the design of future membrane applications.