During the last decade, numerous studies have focused on long-term predictive reactive transport modelling of cement/clay interactions. These simulations have been performed using modelling strategies of growing complexity, e.g. (i) taking more minerals into account, (ii) considering the effect of dissolution/precipitation kinetics versus thermodynamic equilibrium, (iii) refining the spatial discretisation of the models, etc. The present study reviews these simulations in order to identify the main factors affecting numerical results (e.g. mass transport, mesh, selected phases). Simulations are reproduced here with a consistent set of data and input parameters arranged with increasing order of complexity. Only such a standardised approach can allow a proper comparison of numerical results. Modelled reaction pathways (i.e. mineralogical transformations) appear to be independent from the chosen modelling assumptions. Irrespective of the simulated case or the underlying hypotheses, the geochemical transformations remain located very close to the cement/clay interface.