Landslides substantially threaten the socioeconomic status, infrastructure, livestock, and human life in mountainous regions. The West Khasi Hills region is highly susceptible to landslides due to its geological features, slopes, and intense rainfall, which pose significant risks to local communities and infrastructure. Landslide susceptibility map (LSM) delineates regions with a high likelihood of experiencing landslides. The testing and training datasets influence both the development of the LSM and its correctness. This study evaluates the influence of several landslide sampling techniques on the production of LSM. Using the random, grid, and slope-based sample techniques, 191 landslides are split into training and testing datasets. The frequency ratio method uses six conditioning factors and training data to prepare LSM. Results show that 5% of the area is under high and very high susceptibility classes. The slopes within an angle of 20°–29° in the west direction and close to the roads are under high risk. The receiver operating characteristics (ROC) curve’s area under the curve (AUC) value is used in three scenarios to validate the testing landslide points, which have been obtained by sampling methods. The LSM created utilizing slope-based sampling achieved the maximum AUC value of 0.913, followed by grid sampling of 0.895 and random sampling of 0.874. This study provides insights into the application of the sampling method in the preparation of data for LSM studies, and the study has the potential to support infrastructure planning and disaster management in the West Khasi Hills region.