Biological invasion is a major component of global change, and co-invasion of multiple invasive species is becoming increasinglycommon under accelerating globalization, climate warming, and land-use change. Such co-invasions can generate non-additiveimpacts on ecosystems, either exacerbating or mitigating invasion outcomes, yet their ecological consequences and underlyingmechanisms remain poorly understood. We conducted a two-phase greenhouse experiment using Solidago canadensis as a focalinvader and 16 co-occurring invaders to condition soils in monocultures and mixtures, and then tested effects of these soils ona native plant community. We found that non-additive effects of co-invasions on native community biomass were generally neg-ative, with phylogenetically distant co-invaders exerting stronger negative effects. These patterns were largely driven by syner-gistic increases in the richness of putative soil fungal pathogens induced by distantly related co-invaders. A complementary fieldsurvey confirmed these patterns in natural communities, showing that greater phylogenetic distance among co-invaders wasassociated with higher richness of putative soil fungal pathogens and stronger reductions in native plant abundance. Our studyprovides the first empirical evidence that evolutionary relatedness among co-invaders predicts the direction and magnitude oftheir combined impacts via soil microbial pathways. These findings highlight the importance of incorporating evolutionary andbelowground mechanisms into invasion theory, and underscore the urgent need to recognize co-invasion as a key process forpredicting and managing invasion impacts under global change.