Mimicry is a canonical example of adaptive signal design. In principle, what constitutes mimicry is independent of the taxonomic identity of the mimic, the ecological context in which it operates, and the sensory modality through which it is expressed. However, in practice the study of mimicry is inconsistent across research fields, with theoretical and empirical advances often failing to cross taxonomic and sensory divides. We propose a novel conceptual framework whereby mimicry evolves if a receiver perceives the similarity between a mimic and a model and as a result confers a selective benefit onto the mimic. Here, misidentification and/or deception are no longer formal requirements, and mimicry can evolve irrespective of the underlying proximate mechanisms. The centrality of receiver perception in this framework enables us to formally distinguish mimicry from perceptual exploitation and integrate mimicry and multicomponent signalling theory for the first time. In addition, it resolves inconsistencies in our understanding of the role of learning in mimicry evolution, and shows that imperfect mimicry is expected to be the norm. Mimicry remains a key model for understanding signal evolution and cognition, and we recommend the adoption of a unified approach to stimulate future interdisciplinary developments in this fascinating area of research.