A simplified procedure is developed to consider the azimuthal orientation of buildings when estimating seismic risk. Two square-plan reinforced concrete building models are considered as a testbed, one with similar and one with dissimilar properties along the two principal horizontal axes. The fragility of both structures is analysed using a set of ground motion records rotated to multiple incidence angles to develop orientation-dependent fragility functions. It has been observed that, re-orienting all records so that these structures have the same azimuth vis-à-vis the corresponding epicentre leads to significant differences compared to assuming random orientations. Additional results stemming from single-degree-of-freedom oscillators further confirm such findings, showing a dependence to the proximity to the faults and the level of dissimilarity in the principal horizontal axes of the structure. The end results point to a non-negligible bias in assessment studies when a structure's orientation with respect to governing rupture scenarios is not taken into account. It is shown that the median of fragility curves calculated for un-rotated incidence angles can be bias-corrected through shifted by an amount that depends on the azimuthal orientation and level of axes-dissimilarity of structures.