We have simulated exchange of inner-sphere and bulk water molecules for different sizes of Al3+(aq) clusters, Al(H2O)63+ + nH2O for n = 0, 1, 6, or 12, with ab initio and molecular dynamics simulations, in order to understand how robust the ab initio method is for identifying hydrolytic reaction pathways of particular importance to geochemistry. In contrast to many interfacial reactions, this particular elementary reaction is particularly simple and well-constrained by experiment. Nevertheless, we find that a rich array of parallel reaction pathways depend sensitively on the details of the solvation sphere and structure and that larger clusters are not necessarily better. Inner-sphere water exchange in Al3+(aq) may occur through two Langford-Gray dissociative pathways, one in which the incoming and outgoing waters are cis, the other in which they are trans to one another. A large majority of exchanges in the molecular dynamics simulations occurred via the trans mechanism, in contrast to the predictions of the ab initio method. In Al(H2O)63+ + H2O, the cis mechanism has a transition state of 84.3 kJ/mol, which is in good agreement with previous experimental and ab initio results, while the trans mechanism has only a saddle point with two negative frequencies, not a transition state, at 89.7 kJ/mol. In addition to the exchange mechanisms, dissociation pathways could be identified that were considerably lower in energy than experiment and varied considerably between 60 and 100 kJ/mol, depending on the particular geometry and cluster size, with no clear relation between the two. Ab initio calculations using large clusters with full second coordination spheres (n = 12) were unable to find dissociation or exchange transition states because the network of hydrogen bonds in the second coordination sphere was too rigid to accommodate the outgoing inner-sphere water. Our results indicate that caution should surround ab initio simulation of complicated dynamic processes such as hydrolysis, ion exchange, and interfacial reactions that involve several steps. Dynamic methods of simulation need to accompany static methods such as ab initio calculation, and it is best to consider simulated pathways as hypotheses to be tested experimentally rather than definitive properties of the reaction.