Fluorescence in situ hybridisation (FISH) is a powerful tool for visualising the spatial distribution and organisation of microbial communities in their natural environment. However, traditional FISH is limited by the phylogenetic resolution and abundance of the target rRNA. Furthermore, given the time required to design and optimise new rRNA probes and the discovery rate of new lineages, FISH is not scalable to the predicted diversity of microbial species. To address these limitations, we developed GenomeFISH, a high-throughput, scalable, genome-based FISH approach that can differentiate between strains within complex microbial communities. Fluorescent probes are generated from the genomes of single cells, which are collected from environmental samples through high-throughput single-cell sorting. We show that GenomeFISH distinguished between strains with up to 97% average nucleotide identity in mock communities, and when applied to human faecal samples, visually confirmed the presence of two discrete strains of Agathobacter rectalis, a highly abundant health-associated microorganism. By using the whole genome as a target, with >20,000 probe binding sites per genome copy, GenomeFISH reaches superior signal intensities compared to traditional FISH and visualised members of the Patescibacteria, a phylum that is challenging to image with FISH due to its low number of ribosomes. The ability of GenomeFISH to rapidly generate probes from single cells further enables the visualisation of any uncultured lineage, without a priori knowledge of community composition. Thus, it is envisaged that GenomeFISH will become the gold standard in the visualisation of complex microbial systems and its application will provide unprecedented insight into the ecology of these communities.