Box–Behnken (BB) design of response surface methodology (RSM) was effectively applied to optimize fabrication conditions of modified poly(vinyl alcohol) (PVA) and chitosan (CS) blended pervaporation (PV) membranes. The PVA/CS membranes were crosslinked either by chemical reaction with glutaraldehyde (GA) or by heat‐treating at different temperatures. The main objectives were to determine the optimal levels of fabricating parameters and also to investigate interactions among the variables. CS content in the blended membranes, concentration of crosslinking agent and heat‐treating temperature were the fabrication parameters, the main effects and interaction effects of which on membrane structure and PV performance toward isopropanol (IPA)/water dehydration were investigated, and for which regression models were established. The modified PVA/CS blended membranes were characterized by means of scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR) as well as X‐ray diffraction (XRD). It was found that the CS content is the most significant factor influencing flux and separation factor among the three studied variables and the experimental results are in excellent accordance with predicted values from the developed RSM regression models. The RSM results indicated that under preparation conditions of 80 wt % CS in the blended membrane, 0.58 wt % GA concentration, and 77 °C heat‐treating temperature, the maximum separation factor of 5222.8 and the normalized flux of 9.407 kg µm/m2h can be acquired with feed content of 85 wt % IPA at 25 °C, showing that the prepared membrane is highly efficient for PV dehydration of IPA. The models were satisfactorily validated against experimental data. Furthermore, the optimum membrane presents excellent separation performance at different feed compositions and temperatures. © 2016 Wiley Periodicals, Inc. J. Appl. Polym. Sci. 2017, 134, 44587.