Many aspects affect the accuracy of the geographical information system- and statistically-based susceptibility maps. These aspects can be divided into four categories related to: (α) study area, (β) scale, (γ) input data, and (δ) methods, used for susceptibility calculation, landslide representation and map visualisation. Most of these aspects have an impact on the others and each of them is very complex. Due to the complexity of this issue, the article is limited to two aspects, i.e. scale and one of the crucial causal factors, namely geology. The selection of the research area is a key aspect in susceptibility mapping. The research area can influence the choice of scale (pixel size), input data, and methods. In the case of the young glacial Lower Vistula and Central Radunia Valleys, small-scale susceptibility maps give an inaccurate conclusion: on the continental scale, the susceptibility of these areas is underpredicted; on the national scale, the susceptibility is usually overpredicted. The geological structure of young glacial river valleys is complex due to the presence of alternating permeable (sands, gravels) and non-permeable (clays, silts) sediments in the profile, which affects water circulation on the slope and landslide activation. According to the prediction rate method for the Lower Vistula Valley and the classification and regression trees method for the Central Radunia Valley, the maps of surface deposits were of little relevance to landslide susceptibility prediction. However, the inclusion of a quasi-three-dimensional geological model—a typical lithologic sequence layer in the susceptibility mapping of the Vistula Valley, shows that it is one of the most important factors.