Abstract Biological integrity is continuously threatened by human activity. Impacts to biological health may be inferred directly (i.e. bioassessment), or estimated indirectly using rapidly collected field data or computer-based data layer analysis. This study tests whether surrounding land cover alteration (crops, pasture, and development in 200-m buffer) can predict and classify wetland biological health as represented by floristic quality (FQ) and taxonomic distinctness (TD). From surveys of 106 non-forested wetlands in Oklahoma during 2012–2015, we found a limited ability for land use to estimate site-specific FQ and TD values. Although the combined land uses were not a strong predictor overall, pasture land use by itself provided a weak to moderate surrogate when precipitation and ecoregion were taken into account. Floristic quality metrics and average TD decreased with increasing pasture in the 200-m buffer. Land use performed fairly well in classifying sites as biologically intact or not different from a least-altered state (5% FQ and 9% TD misclassification rate), but performed poorly in classifying sites as biologically impaired or deviating from a least-altered state (36% FQ and 24% TD misclassification rate). Importantly, classification analysis suggested that limited land use may separate higher FQ from moderate and lower FQ, whereas more intensive land use may separate lower average TD from moderate and higher average TD. This suggests that both measures (FQ and average TD) are needed for a more complete classification of wetland biological response to anthropogenic disturbance. Despite growing interest in relatively rapid or low-cost methods of wetland assessment, our results do not support using such methods in lieu of direct bioassessment.