This paper proposes a new, computationally efficient, approach to modelling the anisotropy of surface charge on kaolinite particles in particle-scale simulations. We represent each kaolinite particle as a flat ellipsoid and calculate the total interaction energy/force between two ellipsoids as a weighted sum of face-to-face, edge-to-edge, face-to-edge, and edge-to-face interactions. The weightings of these interactions are smooth functions of the relative orientations of the particles. This model was employed in coarse-grained molecular dynamics simulations of virtual samples under isotropic compression to investigate the influence of the pore water pH on microfabric formation and the mechanical behaviour of kaolinite. The simulations effectively captured the influence of pore water pH on microfabric formation and compressive behaviour, as previously reported in experimental research. The virtual sample with a pore water pH of 4 (acidic pore water) formed an aggregated and flocculated microfabric and exhibited higher compressibility during isotropic compression. In contrast, the virtual sample with a pore water pH of 8 (alkaline pore water) formed a dispersed and deflocculated microfabric and showed lower compressibility. The efficacy of the model is demonstrated here through coarse-grained molecular dynamics simulations, but it could also be implemented in a discrete element method code.