Species richness is one of the commonest measures of biodiversity, and is a basis for analyses at multiple scales. Data quality may affect estimations of species richness, but most broad‐scale studies do not take sampling biases into account. We analysed reptile richness on islands that have received different sampling efforts, and assessed how inventory completeness affects the results of ecogeographical analyses. We also used simulations to evaluate under what circumstances insufficient sampling can bias the outcome of biodiversity analyses.

 Mediterranean islands.

We gathered data on reptile richness from 974 islands, assuming better sampling in islands with specific inventories. We used Moran's eigenvector mapping to analyse the factors that determine whether an island has been surveyed, and to identify the relationships between reptile richness, geographical parameters and anthropic parameters. We simulated islands, mimicking patterns of true data, and sampled them with varying effort. Simulated richness was analysed using the same approach used for real‐world data.

The probability that islands were sampled for reptiles was higher in large, human‐populated islands. The relationship between human impact and reptile richness was negative in well‐surveyed islands, but positive in islands that had not been systematically surveyed, because densely populated and accessible islands receive better sampling. In simulations, analyses successfully retrieved the relationships between species richness and human presence only if the average species detection probability was ≥75%. Poorer sampling resulted in biased regression results.

 Human activities may strongly affect biodiversity, but human presence and accessibility improve sampling effort and thus the quality of biodiversity information. Therefore, regressing known species richness on parameters representing human presence may result in apparent positive relationships. These two facets of human presence (positive on biodiversity knowledge, negative on actual biodiversity) represent a major challenge for ecogeographical studies, as not taking them into account would bias analyses and underestimate human impact.