The development of early warning indicators that identify ecosystem stress is a priority for improving ecosystem management. As microbial communities respond rapidly to environmental disturbance, monitoring their composition could prove one such early indicator of environmental stress. We combined 16S rRNA gene sequencing of the seagrass root microbiome of Halophila ovalis with seagrass health metrics (biomass, productivity and Fsulphide) to develop microbial indicators for seagrass condition across the Swan-Canning Estuary and the Leschenault Estuary (south-west Western Australia); the former had experienced an unseasonal rainfall event leading to declines in seagrass health. Microbial indicators detected sites of potential stress that other seagrass health metrics failed to detect. Genera that were more abundant in 'healthy' seagrasses included putative methylotrophic bacteria (e.g. Methylotenera and Methylophaga), iron cycling bacteria (e.g. Deferrisoma and Geothermobacter) and N2 fixing bacteria (e.g. Rhizobium). Conversely, genera that were more abundant in 'stressed' seagrasses were dominated by putative sulphur-cycling bacteria, both sulphide-oxidising (e.g. Candidatus Thiodiazotropha and Candidatus Electrothrix) and sulphate-reducing (e.g. SEEP-SRB1, Desulfomonile and Desulfonema). The sensitivity of the microbial indicators developed here highlights their potential to be further developed for use in adaptive seagrass management, and emphasises their capacity to be effective early warning indicators of stress.