The continued emergence of zoonotic and epizootic diseases among wild animals is a recurrent and often intractable threat to human and animal health, threatening pandemic preparedness 1,2 . Predicting transmission pathways in multi-host systems, like wild birds, where ecological and behavioural diversity interact 3-5, is a particular challenge to surveillance 6-8. As a result, study of disease in such multi-species systems is typically reactive to outbreaks, limiting our ability to predict spatial and temporal patterns of spread. Here, we integrated a fine-scale social network of wild bird co-occurrence with phylogenetic data from High Pathogenicity Avian Influenza Viruses (HPAIVs) to show that a zoonotic disease system can be tracked through a multi-species network, using social links as proxies for transmission probability 9. Links in the network predicted genetic similarity between viral genomes, consistent with HPAIV transmission among species that co-occur in the wild. Species connection strength, independent of abundance, also predicted viral genetic similarity. Our study is the first empirical demonstration of social network structure predicting viral genetic similarity, a proxy for pathogen spread, through a multi-species system, with consequences for surveillance and prioritization of zoonotic and epizootic disease.