Landslides are dangerous events that threaten both human life and property. This research presents a case study of Kaghan valley catchment, which is an area of frequent landslide activity. We aim to present a comparison between the bivariate Landslide Numerical Risk Factor (LNRF), Statistical Index (SI) and Information Value (InfV) statistical models to evaluate landslide susceptibility. A total of 1556 landslides were identified using earlier reports, field surveys, and GOOGLE Earth imagery. The abundance of landslides is primarily controlled by acute deformation caused by a major thrust fault system and proximity to Hazara Kashmir Syntaxis (HKS). A landslide inventory was randomly partitioned into two datasets. 70% (1106) of landslides were used as a training phase of the models, whereas 30% (450) as validation of the three models. A spatial database of 11 conditioning factors was produced consisting of slope, aspect, elevation, lithology, land use, Topographic Wetness Index (TWI), rainfall, Stream Power Index (SPI), distance to faults, rivers and streams. All the landslide susceptibility assessment parameters were obtained from different sources and different landslide susceptibility maps were prepared on the GIS software. Performance of the three models was validated through the Receiver operator Characteristics (ROC) through success and prediction rate curves. Results show that the area under the ROC curve (AUC) for InfV, LNRF and SI models are 70.95, 83.99 and 67.56 for success rate curves and 70.75, 83.99 and 67.85 for prediction rate curves, respectively. LNRF having the highest AUC value proved to be superior for generating regional scale landslide susceptibility maps.