Metabarcoding can generate large numbers of georeferenced occurrence data from bulk samples at low cost. Its integration into the practice of agricultural invertebrate biomonitoring currently lacks both standard methods and example datasets that allow the identification of potential challenges and uncertainties. For this study, we gathered metabarcoding data of terrestrial arthropods from Malaise trap samples across sites in southern Ontario, spanning a gradient from high production, intensely farmed areas to alternative land use farms with varying amounts of natural restoration of marginal lands. The result is one of the largest datasets available for comparison of how agricultural practices influence arthropod biodiversity.