A model is proposed for the heterolytic dissociation of covalent bonds at the surface of perfect or defective oxide surfaces essentially based on the analysis of the electrostatic potential and its gradient at the bare surface. It permits us to obtain in a semiquantitative way the equilibrium geometry of the fragments and the reaction energy, and gives clues for the identification of the reaction path and for a rough estimate of the related activation energy. The predictive ability of the model has been verified by performing a number of calculations to simulate H-2 dissociation at various defects at MgO and CaO: the isolated oxygen vacancy, the divacancy at the (100) face, the infinite edge and the divacancy at the edge. All calculations have been performed at an ab initio Hartree-Fock level of approximation, using the CRYSTAL program for the periodic structures and the EMBED program for the local defects. Generally satisfactory agreement is found between the model predictions and the results of actual calculations. The model could be useful for predicting with low cost computations if and how a local structure at a defective oxide surface is capable or not of heterolytically dissociating strong covalent bonds.