The predictive power of species distribution models (SDMs) varies substantially among species depending on their ecological and life-history traits, but which of these traits are the most relevant and how they influence species ‘predictability’ remains an area of debate. Here, we address these questions in bryophytes. SDMs employing macroclimatic, topographic and edaphic predictors were calibrated for 411 species in Europe and externally evaluated using an independent dataset. Regression models were implemented to determine whether species characteristics, including life-history traits, ecological preference and niche breadth, determine the accuracy of SDMs. Variation in SDM accuracy among species was significantly explained by species characteris-tics, supporting the hypothesis that the strength of species–environment correlations is affected by characteristics of the species themselves. The percent variance of SDM accuracy explained by species traits, however, substantially varied between 9 and 57% depending on the evaluation metrics used. The lower correlation observed between species traits and MaxKappa and the Boyce index than with area under the curve (AUC) and MaxTSS suggests that the former are less suitable than the latter for deter-mining species ‘predictability’ based on their traits. SDM accuracy decreased from species restricted to pristine habitats to species thriving in eutrophic habitats with high levels of human disturbance. The widespread distribution of man-made habitats in fact opens the door for the spread of now ubiquitous species, even in environments that would primarily not be suitable for them. Such species, likely to occur anywhere, reach very high to full occupancy rates, thereby decreasing the accuracy of models aiming at predicting their distributions. The fact that AUC and MaxTSS were higher for species from pristine habitats is important in a conservation context, as ubiquitous species from eutrophic, disturbed environments are precisely the ones of lower conservation relevance.