Different approaches exist to incorporate faults in reservoir-scale geomechanical models. Challenges are the proper representation of the fault geometry as well as the small-scale variations in the internal architecture and the mechanical properties of the fault zone regarding the typical size of such models, i.e., kilometres to tens of kilometres. The present study utilizes a simple generic fault zone model to compare three different possibilities commonly used to represent faults in finite element reservoir models. Two differ in the basic grid geometry and the arrangement of mechanically weak fault zone elements, respectively. The third uses a discontinuous grid and contact elements to represent the fault. Modelling results show remarkable differences in the calculated stress and strain patterns. The relatively strongest perturbations result for a continuous curvilinear grid adapted to the fault geometry. In contrast, the fault implementation has the least impact on local stresses and strains if it is represented as a stair-step structure contained in a rectangular grid. The use of contact elements has an intermediate effect. Modelling results are used to infer some general recommendations concerning the appropriate approach of representing faults in a numerical-geomechanical reservoir models depending on fault geometry, model scale and scope of interest.