This review discusses chiral self-sorting-the process of choosing an interaction partner with a given chirality from a complex mixture of many possible racemic partners. Chiral self-sorting (also known as chiral self-recognition or chiral self-discrimination) is fundamental for creating functional structures in nature and in the world of chemistry because interactions between molecules of the same or the opposite chirality are characterized by different interaction energies and intrinsically different resulting structures. However, due to the similarity between recognition sites of enantiomers and common conformational lability, high fidelity homochiral or heterochiral self-sorting poses a substantial challenge. Chiral self-sorting occurs among natural and synthetic molecules that leads to the amplification of discrete species. The review covers a variety of complex self-assembled structures ranging from aggregates made of natural and racemic peptides and DNA, through artificial functional receptors, macrocyles, and cages to catalytically active metal complexes and helix mimics. The examples involve a plethora of reversible interactions: electrostatic interactions, π-π stacking, hydrogen bonds, coordination bonds, and dynamic covalent bonds. A generalized view of the examples collected from different fields allows us to suggest suitable geometric models that enable a rationalization of the observed experimental preferences and establishment of the rules that can facilitate further design.