Cultural ecosystem services (CESs) intangibly influence many aspects of public daily life, and when evaluating them, it is difficult to obtain perception data. Spatial value transfer predicts the ecosystem service value of policy sites based on the spatial associations of study sites with services value. This approach has the potential to produce an evaluation without perception data. Previous studies have preliminarily researched the validity and effects of using spatial value transfer to evaluate CESs, but, there are doubts about the transfer performance of indicators, differences among various CES types, and the impact of the physical environment. This study examines the above key points of spatial value transfer in depth, with urban parks as the study area to assess three types of CES values that are closely related to human well-being: aesthetic, historical and recreational value. Two popular indicators, area under the curve statistics and transfer error rate, are used to evaluate transfer performance. The results reveal that historical values have the best transfer performance, and transfer error is lower in transfers from small locations to large locations with environmental variable combinations that include road network, water coverage and attraction distribution. The transfer error rate indicator could not only assess the overall transfer performance but also generate maps to reflect the distribution and value of transfer error. We suggest that more knowledge of minor differences in physical factors and developing specific transfer coefficients for them are necessary to enhance the accuracy of CES evaluation by spatial value transfer.