Climate-driven range shifts may cause local extinctions, while the accompanying loss of biotic interactions may trigger secondary coextinctions. At the same time, climate change may facilitate colonizations from regional source pools, balancing out local species loss. At present, how these extinction–coextinction–colonization dynamics affect biological communities under climate change is poorly understood. Using 84 communities of interacting plants and hummingbirds, we simulated patterns in climate-driven extinctions, coextinctions and colonizations under future climate change scenarios. Our simulations showed clear geographic discrepancies in the communities’ vulnerability to climate change. Andean communities were the least affected by future climate change, as they experienced few climate-driven extinctions and coextinctions while having the highest colonization potential. In North America and lowland South America, communities had many climate-driven extinctions and few colonization events. Meanwhile, the pattern of coextinction was highly dependent on the configuration of networks formed by interacting hummingbirds and plants. Notably, North American communities experienced proportionally fewer coextinctions than other regions because climate-driven extinctions here primarily affected species with peripheral network roles. Moreover, coextinctions generally decreased in communities where species have few overlapping interactions, that is, communities with more complementary specialized and modular networks. Together, these results highlight that we should not expect colonizations to adequately balance out local extinctions in the most vulnerable ecoregions.