Ground motion, one of the key parameters in controlling coseismic landslides, has long been treated as a scalar in the regional earthquake landslide analysis, which is apparently not sufficient for understanding coseismic landslide hazards. Using the 2017 Jiuzhaigou M(W)6.5 earthquake and its observed coseismic landslides as a case study, we aim to explore the relationship between coseismic landslides and ground motion with significant spatial variation, as computed by physics- and scenario-based Neo-deterministic Seismic Hazard Assessment (NDSHA). We calculate 84 peak ground motion parameters for PGA, PGV and PGD related to different direction and frequencies, and statistically analyze their correlation with landslide characteristic parameters. Our findings reveal: (i) the absolute values of the correlation coefficients between the Radial and North-South components of ground motion and the landslide parameters are the largest; (ii) the parameters of low-frequency waves show the highest proportion and correlation coefficients for positive correlations with point density, area density, and landslide area, while parameters of high-frequency waves correlate better with the landslide mobility parameter; (iii) The group of PGA and PGD parameters display better correlations with the four landslide parameters. (iv) Our results improve the accuracy of coseismic landslide susceptibility by up to 8.4 % compared to the widely used PGA ShakeMap from the USGS. This study indicates that no single peak ground motion parameter shows a leading correlation with all characteristic parameters of landslide, and different peak ground motion parameters are needed for different landslide parameters for improving the accuracy and applicability of the regional coseismic landslide assessment.