We have conducted experiments to evaluate the vapour–liquid fractionation of Mo(VI) in the system MoO₃–NH₃–H₂O at 300–370°C and saturated vapour pressure, using a two-chamber autoclave that allows separate trapping of the vapour and liquid. The measured total Mo concentrations in each phase were used to calculate a distribution coefficient, K_D^V/L, which increases as the density of the vapour approaches that of the liquid, and is greater than one for pH⩽4. Molybdenum speciation in the vapour is described by a single complex, MoO₃H₂O. By contrast, thermodynamic modeling of the distribution of Mo species in the liquid indicates that bimolybdate (HMoO₄⁻) is the dominant aqueous species at the conditions of our experiments, and that molybdate (MoO₄²⁻) and molybdic acid (H₂MoO₄⁰) are present in smaller quantities. As vapour–liquid fractionation occurs between neutral species, it is governed by the reaction H₂MoO₄⁰(aq)=MoO₃·H₂O(g). Fractionation is therefore controlled by the concentration of H₂MoO₄⁰ in the liquid, which increases with increasing temperature and decreasing pH. Owing to the pH dependence of K_D^V/L, it cannot be used to describe Mo fractionation in aqueous vapour–liquid systems with compositions different than those of this study. We have therefore calculated a composition-independent (Henry’s Law) constant, K_H^V/L, for each experimental point, using the measured total Mo concentration in the vapour and the modeled concentration of H₂MoO₄⁰ in the liquid. This constant may be applied to aqueous vapour–liquid systems of known liquid composition to estimate the concentration of Mo in a vapour for which little chemical information is available, and thereby supplement the available fractionation data for natural porphyry-forming systems. The results of this study demonstrate that at conditions typical of natural porphyry ore-forming systems, a significant amount of molybdenum fractionates into the vapour over the liquid, and the vapour may transport quantities of Mo in excess of that in the liquid at pH conditions below those of the muscovite–microcline reaction boundary.