Historical biogeography analyses, used to infer spatial dynamics of organisms through their evolutionary history, often consider the distribution of higher taxa (like that of a genus, subfamily, family, etc.) despite only including a subsample of their representatives. When tips of the tree have the ranges of multiple species instead of individual species, the outcomes of all relevant methods and softwares are conceptually wrong. This is because they are designed for species-level phylogenies and can thus lead to several biases: widespread ancestors, high uncertainties and wrong estimation of dispersal probabilities. Here we present an approach that aims at minimizing these biases through biogeographical analyses based on multiple randomizations of the possible ranges for a given group. We computed historical biogeography analyses based on simulated random trees and infer the ancestral ranges from the group’s ranges considering three settings: (i) classical, considering all areas represented within the group; (ii) randomizing possible ranges for the group; and (iii) as (ii), but weighting according to the actual ranges of species within the group. In addition, we used these three approaches on a genus-level phylogeny of wild silkmoths (Lepidoptera, Saturniidae). Our results from simulated phylogenies revealed that randomization approaches are more precise than classical analyses. They also helped resolve the complex biogeographical dynamics of Saturniidae. We emphasize here the biases induced when considering multiple species ranges at the tips of a phylogeny, and recommend the use of a randomization approach when inferring biogeographical