Forbidden links are defined as pairwise interactions that are prevented by the biological traits of the species. We focus here on the neglected importance of intraspecific trait variation in the forbidden link concept. We show how intraspecific trait variability at different spatiotemporal scales, and through ontogeny, reduces the expected prevalence of forbidden interactions. We also highlight how behavior can foster interactions that, from traits, would be predicted to be forbidden. We therefore discuss the drawbacks of frameworks recently developed to infer biotic interactions using available trait data (mean values). Mispredictions can have disproportionate effects on inferences about community dynamics. Thus, we suggest including intraspecific variability in trait-based models and using them to guide the sampling of real interactions in the field for validation.