Species introductions often bring together genetically divergent source populations, resulting in genetic admixture. This geographic reshuffling of diversity has the potential to generate favorable new genetic combinations, facilitating the establishment and invasive spread of introduced populations. Observational support for the superior performance of admixed introductions has been mixed, however, and the broad importance of admixture to invasion questioned. Under most underlying mechanisms, admixture’s benefits should be expected to increase with greater divergence among and lower genetic diversity within source populations, though these effects have not been quantified in invaders. We experimentally crossed source populations differing in divergence in the invasive plant Centaurea solstitialis. Crosses resulted in many positive (heterotic) interactions, but fitness benefits declined and were ultimately negative at high source divergence, with patterns suggesting cyto-nuclear epistasis. We explored the literature to assess whether such negative epistatic interactions might be impeding admixture at high source population divergence. Admixed introductions reported for plants came from sources with a wide range of genetic variation, but were disproportionately absent where there was high genetic divergence among native populations. We conclude that while admixture is common in species introductions and often happens under conditions expected to be beneficial to invaders, these conditions may be constrained by predictable negative genetic interactions, potentially explaining conflicting evidence for admixture's benefits to invasion.