Biodiversity conservation measures designed to ensure ecological connectivity depend on the reliable modeling of species movements. Least cost path modeling makes it possible to identify the most likely dispersal paths within a landscape and provide two items of ecological relevance: (i) the spatial location of these least cost paths (LCPs) and (ii) the accumulated cost along them ('cost distance', CD). This spatial analysis requires that cost values be assigned to every type of land cover. The sensitivity of both LCPs and CDs to the cost scenarios has not been comprehensively assessed across realistic landscapes and diverging cost scenarios. We therefore assessed it in diverse landscapes sampled over metropolitan France and with widely diverging cost scenarios. The spatial overlap of the LCPs was more sensitive to the cost scenario than the CD values were. Besides, highly correlated CD matrices could derive from very different cost scenarios. Although the range of the cost values and the properties of each cost scenario significantly influenced the outputs of LCP modeling, landscape composition and configuration variables also explained their variations. Accordingly we provide guidelines for the use of LCP modeling in ecological studies and conservation planning.