Plant diversity can reduce the risk of plant disease, but positive, and neutral effects have also been reported. These contrasting relationships suggest that plant community composition, rather than diversity per se, affects disease risk. Here, we investigated how the diversity and composition of plant communities drive root-associated pathogen accumulation belowground. In a temperate grassland biodiversity experiment, containing 16 plant species (forbs and grasses), we determined the abundance of root-associated fungal pathogens in individual plant species growing in monocultures and in four-species mixtures through Illumina MiSeq amplicon sequencing. In the plant monocultures, we identified three major fungal pathogens that differed in host range: Paraphoma chrysanthemicola, associated with roots of forb species of the Asteraceae family, Slopeiomyces cylindrosporus, associated with grass species, and Rhizoctonia solani, associated with multiple forb and grass species. In mixtures, there was no significant reduction in relative abundance of these pathogens in their host species as compared to monocultures. However, in mixtures, there was a significant increase in relative abundance of each pathogen in several non-host and host plant species. Across mixtures, plant community composition affected pathogen relative abundance in individual plant species. This effect was driven by the presence of a particular neighbouring plant species (depending on the pathogen), rather than functional group composition (i.e. grass/forb ratio) or averaged pathogen pressure (based on monocultures) of all neighbours. Specifically, the presence of neighbour host species Achillea millefolium significantly increased P. chrysanthemicola, but decreased R. solani relative abundance in several host and non-host plant species in mixtures. Synthesis. Our results indicate that interactions between different plant species—both host and non-hosts—and fungal pathogens underlie the effects of plant diversity on root pathogen abundance. Non-host species may act as pathogen reservoirs in diverse plant communities, as they harboured certain pathogens in mixtures, but not in monocultures. Additionally, particular host species can strongly affect pathogen abundance in other (host and non-host) plant species in plant mixtures, suggesting clear effects of species identity in the diversity–disease relationship. Below-ground disease risk thus depends on plant community composition rather than diversity per se, via specific interactions between plant species and their root-associated pathogens.