Acoustic seabed classification (ASC) is an important method for understanding landscape-level physical and biological patterns in the aquatic environment. Bottom habitats in the Laurentian Great Lakes are poorly mapped to date, and will require a variety of contributors and data sources to complete. We repurposed a long-term split-beam echosounder dataset gathered for purposes of fisheries assessment to estimate lakebed properties utilizing unsupervised classification of echo return data. We interpreted first echo properties representing lakebed hardness and roughness to define and map three statistically supported lakebed classes revealed through cluster analysis. Our results indicate coherent and logical class boundaries and suggest that the dataset has promise for expanded use in ASC. To improve inferences using repeated measures, future work should focus on collecting ground truth information for areas previously surveyed and on collecting concurrent ground truth information when sampling acoustic data moving forward.