The kinetochore is a dynamic multi-protein assembly that forms on each sister chromatid and interacts with microtubules of the mitotic spindle to drive chromosome segregation. In animals, kinetochores without attached microtubules expand their outermost layer into crescent and ring shapes to promote microtubule capture and spindle assembly checkpoint (SAC) signalling. Kinetochore expansion is an example of protein co-polymerization, but the mechanism is not understood. Here, we present evidence that kinetochore expansion is driven by oligomerization of the Rod-Zw10-Zwilch (RZZ) complex, an outer kinetochore component that recruits the motor dynein and the SAC proteins Mad1-Mad2. Depletion of ROD in human cells suppresses kinetochore expansion, as does depletion of Spindly, the adaptor that connects RZZ to dynein, while dynein itself is dispensable. Expansion is also suppressed by mutating ZWILCH residues implicated in Spindly binding. Conversely, supplying cells with excess ROD facilitates kinetochore expansion under otherwise prohibitive conditions. Using the C. elegans early embryo, we demonstrate that ROD-1 has a concentration-dependent propensity for oligomerizing into μm-scale filaments, and we identify the ROD-1 β-propeller as a key regulator of self-assembly. Finally, we show that a minimal ROD-1-Zw10 complex efficiently oligomerizes into filaments in vitro. Our results suggest that RZZ's capacity for oligomerization is harnessed by kinetochores to assemble the expanded outermost domain, in which RZZ filaments serve as recruitment platforms for SAC components and microtubule-binding proteins. Thus, we propose that RZZ self-assembly into filaments underlies the adaptive change in kinetochore size that contributes to chromosome segregation fidelity.