We introduce a procedure for selecting and ranking of ground-motion prediction equations (GMPEs) that can be useful for regional or site-specific probabilistic seismic hazard assessment (PSHA). The methodology is called Euclidean distance-based ranking (EDR) as it modifies the Euclidean distance (DE) concept for ranking of GMPEs under a given set of observed data. DE is similar to the residual analysis concept; its modified form, as discussed in this paper, can efficiently serve for ranking the candidate GMPEs. The proposed procedure separately considers ground-motion uncertainty (i.e., aleatory variability addressed by the standard deviation) and the bias between the observed data and median estimations of candidate GMPEs (i.e., model bias). Indices computed from the consideration of aleatory variability and model bias or their combination can rank GMPEs to design GMPE logic trees that can serve for site-specific or regional PSHA studies. We discussed these features through a case study and ranked a suite of GMPEs under a specific ground-motion database. The case study indicated that separate consideration of ground-motion uncertainty (aleatory variability) and model bias or their combination can change the ranking of GMPEs, which also showed that the ground-motion models having simpler functional forms generally rank at the top of the list. We believe that the proposed method can be a useful tool to improve the decision-making process while identifying the most proper GMPEs according to the specific objectives of PSHA.