Ground motion records are generally required to develop fragility curves which are utilized in seismic loss estimation studies for earthquake prone zones. These records could be 'real', i.e. recorded acceleration time series or 'simulated' records consistent with the regional seismicity and produced using alternative simulation techniques. This study mainly concentrates on development of fragility curves for masonry buildings located in Erzincan (eastern Turkey) using 'simulated' ground motion records and evaluates the suitability of these fragilities by comparing with the curves developed utilizing 'real' records. To generate a set of scenario earthquakes by using regional seismicity parameters of Erzincan, stochastic finite-fault methodology has been employed as the simulation technique. The 'real' records, on the other hand, are selected from Pacific Earthquake Engineering Research Center's NGA-West2 ground motion database and chosen to be compatible with the seismological characteristics of the region of interest. The records are imposed on a set of masonry buildings (with 1-, 2-, 3-stories and 3 different quality classes) representing the building stock in Erzincan. To assess the dynamic responses of the structures, nonlinear time history analyses utilizing equivalent single-degree-of-freedom systems are performed on OpenSees platform. Afterwards, fragility curves for three limit states (Immediate Occupancy, Life Safety and Collapse Prevention) are generated. This study also investigates the sensitivity of these fragility curves to ground motion variability and two alternative fragility curve generation methods. (c) 2017 The Authors. Published by Elsevier Ltd.