The conservation of biodiversity has gained prominence in ecological research for the last decades. Conservation actions require a measure of biodiversity such as species richness, but its assessment is very difficult, even for small areas and therefore the search for surrogates (i.e. indicators) of biodiversity has emerged as an active research topic. We investigated the relationships between butterfly species richness and landscape structure and composition in two pine plantation sites in Southwest France. We assessed the correlation between butterfly species richness and a set of 15 landscape metrics computed for 18 land-uses at 10 different spatial scales. Spatial scales were accounted for by computing landscape metrics for circular buffers with radius ranging from 100 to 1000 m. The joint use of the Partial Least Squares Regression (PLSR) and a stepwise regression procedure revealed strong correlations between butterfly species richness and various landscape metrics in both study sites. The selected landscape metrics differed from one site to another and mostly involved measures of landscape fragmentation. We found a very strong effect of the spatial scale of investigation upon the perception of the landscape–butterfly richness relationship. Our main conclusions are that (i) certain landscape attributes can potentially serve as indicators for butterfly species richness at the landscape scale; (ii) future indicators of biodiversity based on landscape features should consider various spatial scales.