Geotechnical engineers are well aware that the particle surface chemistry and the pore fluid composition can significantly influence the mechanical behaviour of clay. Reference is often made to the Derjaguin-Landau-Vervey-Overbeek (DLVO) theory, which enables the electrochemical interactions between charged particles to be estimated. Hitherto, the absence of an effective framework for particle-scale simulation of clay has inhibited a direct link between these electrochemical interactions and clay behaviour (e.g. load:deformation response) or fabric (i.e. the development of a disperse or flocculated fabric). Ebrahimi [1] demonstrated the viability of using molecular dynamics simulations where the clay grains are simulated as ellipsoidal particles whose interactions are described by an analytical expression called the Gay-Berne (GB) potential. While promising when compared to other approaches documented in the literature, Ebrahimi's work considered only a single clay mineralogy and did not explicitly account for the pore fluid composition. This paper considers the use of the Gay-Berne potential in particle-scale modelling of clay from a more general perspective. Calibration of the GB model parameters to predict kaolinite particle interactions reveals a lack of generality in Ebrahimi's approach. The Gay-Berne potential cannot simulate situations in which attractive and repulsive interactions co-exist, which lead to the classical “cardhouse” fabric, as is the case of kaolinite particles interacting via an acidic pore fluid.