Summary Generally, in classical reservoir studies, the geomechanical behavior of the porous medium is taken into account by the rock compressibility. Inside the reservoir simulator, the rock compressibility is assumed to be constant or to vary with the pressure of the oil phase. It induces some changes in the porosity field. During the depletion phase or the cold-water injection of high-pressure/high-temperature (HP/HT) reservoirs, the stress state in and around a reservoir can change dramatically. This process might result in rock movements such as compaction, induced fracturing, and enhancement of natural fractures and/or fault activation, which continuously modify the reservoir properties such as the permeabilities and the fault transmissibilities. Modifications of such parameters strongly affect the flow pattern in the reservoir and ultimately the recovery factor. To capture the link between flow and in-situ stresses, it becomes essential to conduct coupled reservoir-geomechanical simulations. This paper compares the use of five types of approach for the reservoir simulations: A classical approach with rock compressibility using only a reservoir simulator. A loose coupled approach between a reservoir simulator (finite volumes) and a geomechanical simulator (finite elements). At given user-defined steps, the hydrocarbon pressures calculated by the reservoir simulator are transmitted to the geomechanical tool, which computes the actual stresses and feeds back iteratively the modifications of the petrophysical properties (porosities and permeabilities) to the reservoir simulator. A one-way coupling: this approach is a simplification of the loose coupled approach in that the modifications are not fed back to the reservoir simulator. A simplified approach using permeability and porosity multipliers inside a reservoir simulator. These multipliers are user-defined curves and vary with the pressure of the oil phase. This approach uses only a reservoir simulator. A coupled approach in which the structural and the flow unknowns (displacement, pressure, and saturations) are solved simultaneously. These approaches are compared for two validation cases and two field cases described in the following.