Summary

Background: Many studies have tried to estimate the number of undescribed species based on the known diversity. These estimates often rely on extrapolation based on data from a limited number of species. Although statistical methods provide accurate inference when generalizing from a random sample, their predictions will be biased when based on a non-random sample unless the sampling process is explicitly accounted for. Problem: In this paper, we argue that this is a fundamental issue in many estimates of unrecorded biodiversity. We show that the sample of species used in biodiversity extrapolation represents disproportionately common and abundant taxa, which leads to an overestimation of extrapolated diversity. We discuss this issue in the context of three specific cases: estimates of plant-associated insect diversity, estimates of parasite diversity and estimates of cryptic species diversity. Implications: For the example of plant-associated insects, we provide an estimate of the magnitude of the error, but insufficient data are currently available to estimate the magnitude of the problem for the other examples. Our findings cast doubt over previous attempts to estimate the number of undescribed species, suggesting that they provide consistent overestimations. © 2026 The Author(s). Global Ecology and Biogeography published by John Wiley & Sons Ltd.