Abstract Invasive alien species are among the most concerning threats to native biodiversity world‐wide, and the level of landscape heterogeneity is considered to affect spatial patterns of their occurrence and spread. However, as previous studies on these associations report contrasting results, the role of landscape heterogeneity on its susceptibility to invasions remains poorly understood. Landscape heterogeneity is usually described by two measures: configuration and composition. Both measures may differently affect invasive species and these impacts may be additionally scale dependent. Nevertheless, their relative contribution to invasion patterns is poorly known. We investigated the effect of two landscape heterogeneity components: configuration (edge density) and composition (number and evenness of land cover types) measured at different spatial scales (from within 0.25 to 5 km of the studied locations) on the local abundance of one of the most invasive alien plant species in Europe, the North American goldenrods (Solidago canadensis and S. gigantea). Using publicly available geospatial environmental data and a novel method based on remote analysis of Google Street View images, we collected and analysed large dataset on goldenrod occurrence along 1,347 roadside transects in agricultural landscapes of Poland. Both the compositional and configurational heterogeneity were positively associated with the local abundance of goldenrods, however, the effect size of these relationships was dependent on spatial scale. While abundance–heterogeneity associations were most pronounced at the largest spatial scale for compositional heterogeneity, the pattern was the opposite for configurational heterogeneity. Synthesis and applications. Landscape heterogeneity is a clear correlate of plant invasion potential, with occurrences of invasive plants generally higher in more heterogeneous landscapes. However, scale dependence of this association means that researchers and practitioners may miss the association if only concentrating on a single spatial scale. While increasing heterogeneity of rural landscapes is widely introduced as a way to promote farmland biodiversity, we show that it may also support invasive plants, and thus conflict with original goals of biodiversity‐oriented strategies. Therefore, we suggest implementing regular management and eradication schemes in most heterogeneous landscapes. Finally, we demonstrate how remote analysis of plant invasions using existing imagery can advance our understanding of invasion biology.