We analyzed data from coastal wetlands across the Laurentian Great Lakes to identify fish assemblage patterns and relationships to habitat, watershed condition, and regional setting. Nonmetric multidimensional scaling (NMDS) ordination of electrofishing catch-per-effort data revealed an overriding geographic and anthropogenic stressor gradient that appeared to structure fish composition via impacts on water clarity and vegetation structure. Wetlands in Lakes Erie and Michigan with agricultural watersheds, turbid water, little submerged vegetation, and a preponderance of generalist, tolerant fishes occupied one end of this gradient, while wetlands in Lake Superior with largely natural watersheds, clear water, abundant submerged vegetation, and diverse fishes occupied the other. Fish composition was also related to wetland morphology, hydrology, exposure, and substrate, but this was only evident within low-disturbance wetlands. Anthropogenic stress appears to homogenize fish composition among wetlands and mask other fish–habitat associations. Because land use is strongly spatially patterned across the Great Lakes and aquatic vegetation is a key habitat element that responds to both biogeography and disturbance, it is difficult to disentangle natural from anthropogenic drivers of coastal wetland fish composition.