This study describes the methodology implemented to establish the ground-motion logic-tree for national probabilistic seismic hazard map of Turkey for shallow active crustal regions. The presented procedure provides quantitative information to guide the hazard experts while establishing the logic tree to capture the epistemic uncertainty in ground-motion characterization. It uses non-data-driven and data-driven testing methods to identify and rank candidate ground-motion prediction equations (GMPEs) under a specific ground-motion database. The candidate GMPEs are subjected to visual inspection and are classified into center, body and range (CBR) spectral estimates for a proper consideration of epistemic uncertainty. The GMPEs classified into CBR are then used in a suite of seismic hazard sensitivity analysis to establish the most suitable GMPE logic-tree whose spectral estimates are not biased by any one of the GMPEs in the logic-tree structure. The sensitivity analysis considers normalized spectral ordinates and is not manipulated by the spectral amplitudes. The proposed procedure is inherited from the relevant studies of the Earthquake Model of the Middle East (EMME; ) regional seismic hazard project. This paper also highlights the similarities and differences in ground-motion characterization between EMME and our approach.