Faults in the brittle crust constitute preexisting weakness zones that can be reactivated depending on their friction, orientation within the local stress field, and stress field magnitude. Analytical approaches to evaluate the potential for fault reactivation are generally based on the assumption that faults are ideal planes characterized by zero thickness and constant friction. However, natural faults are complex structures that typically host thick fault rocks. Here we experimentally investigate the reactivation of gouge-bearing faults and compare the resulting data with theoretical predictions based on analytical models. We simulate preexisting faults by conducting triaxial experiments on sandstone cylinders containing saw-cuts filled with a clay-rich gouge and oriented at different angles, from 30 degrees to 80 degrees, to the maximum principal stress. Our results show the reactivation of preexisting faults when oriented at 30 degrees, 40 degrees, and 50 degrees to the maximum principal stress and the formation of a new fracture for fault orientations higher than 50 degrees. Although these observations are consistent with the fault lock-up predicted by analytical models, the differential stress required for reactivation strongly differs from theoretical predictions. In particular, unfavorable oriented faults appear systematically weaker, especially when a thick gouge layer is present. We infer that the observed weakness relates to the rotation of the stress field within the gouge layer during the documented distributed deformation that precedes unstable fault reactivation. Thus, the assumption of zero-thickness planar fault provides only an upper bound to the stress required for reactivation of misoriented faults, which might result in misleading predictions of fault reactivation.