Landscape linkages are hypothetical objects developed in geographic information systems (GIS) proposed to connect areas of habitat in fragmented landscapes. Although there are specific places and species (primarily ungulates) where hypothetical linkages represent where wildlife move, for the majority of places and species there is very little evidence that these GIS objects represent an ecological reality. Assuming they were used by organisms in nature, linkages could be an important tool for the maintenance of viable wildlife populations and biodiversity conservation. Surprisingly, large-extent connectivity models (e.g., at the US state scale) are generally not based on evidence of wildlife occurrence and the testing of whether or not wild animals follow the pathways created by conservation planners have given mixed results. We evaluated five California linkage models (four statewide and one desert specific) for use by common California mammal, reptile and amphibian species in two ways, using >180,000 wildlife detections over 20 years. First, we focused on roadkill detections for 10 species as a proxy for wildlife movement across landscapes and tested whether linkages were an important predictor for the presence of these detections. The idea in this case is that if wildlife were following linkages, then there should be a concentration of roadkill where roadways intersect linkages. For the species we evaluated we found that the linkage areas were not consistent predictors of conflict on roads and most species generally did not avoid occupancy in areas closer to roads, with the exception of certain large-bodied mammals by individual models. For certain species and linkage model combinations, presence of linkages predicted presence of roadkill, but for most species and linkages, there was no statistically-significant relationship. Linkage areas were also not important predictors for the probability of landscape occupancy for most species. Unless validated using data from wildlife movements, caution should be exercised when using linkage models in land and transportation-based conservation planning. In urban environments, modeled linkages may reflect the critical last places for wildlife to move and should be reserved. However, in agricultural, forested, rangeland, and other mixed-use landscapes where wildlife may move more easily, connectivity as a gradient across the landscape should be the target of conservation, including in land-use and transportation planning. As proposed by L Fahrig (Fahrig, 2018, DOI: 10.1111/geb.12839), pride among conservation scientists may have led to unproven concepts in fragmentation and connectivity analysis being retained regardless of evidence and concluded: When the data do not support the prediction, we need to revise our thinking and change the hypothesis.