We quantified the effect of seed orientation deviations on five prostate seed implant cases at our institution. While keeping their positions fixed, the iodine-125 seeds were assigned orientations sampled from a realistic probability distribution derived from the post-implant radiographs of ten patients. Dose distributions were calculated with both a model that explicitly includes anisotropy (TG43 anisotropy function) and a point source model (TG43 anisotropy factor). Orientation deviations had only a small influence on prostate dose-volume histograms: the 95% confidence intervals on the volumes receiving 100%, 150% and 200% dose were at most +/-0.8%, +/-1.1% and +/-0.6% of the prostate volume, respectively. The dose-volume histograms of anisotropic seed distributions were marginally better than those with isotropic point-source seeds. Anisotropy caused a displacement of cold spots (regions receiving <100% of the prescribed dose) in <1% of the prostate volume. Our results indicate no net benefit to prostate dosimetry in using more isotropic seeds. Furthermore, we propose a new 'weighted anisotropy function' to better account for the effects of anisotropy when seed orientation is unknown. Conceptually, the TG43 anisotropy factor described in AAPM TG43 averages the effect of anisotropy over all solid angles, with the implicit assumption that all seed orientations are equally probable. In prostate implants, however, seeds are preferentially oriented parallel to the needle axis. The proposed weighted anisotropy function incorporates this non-uniform probability.