•Development of aquatic nutrient criteria often rely on relationships with biological indices.•Summarizing taxa to biological indices may drive weak relationships with nutrients.•Nutrient criteria setting should consider relationships between nutrients and individual taxa. Nutrient enrichment is one of the most pervasive impacts on aquatic ecosystems globally. Approaches to establish nutrient criteria that safeguard aquatic ecosystem health are highly variable and, in many instances, criteria are derived from correlations between in-situ nutrient concentrations and biological indices. Summarising entire assemblages with a single index can result in a substantial loss of information and potentially weaker relationships. In this study, we compared the derivation of nutrient criteria using biological indices and those from individual taxa for rivers and streams in New Zealand. Random forest models, including nutrient concentrations, were built to predict two biological indices and individual taxa across New Zealand's river monitoring network. For all acceptable models, the response of the biological indices and individual taxa to increasing Dissolved Inorganic Nitrogen (DIN) and Dissolved Reactive Phosphorus (DRP) were then predicted for every river reach across the nation, and nutrient concentrations that protected 80% of taxa were then identified. Models for the biological indices were poor but were good for most of the taxa, with nutrient concentrations almost always being the most influential factor. To ensure persistence of at least 80% of the taxa within a river reach, we estimated that DIN (Dissolved Inorganic Nitrogen) concentrations would need to be below 0.57–1.32 mg/L, and DRP (Dissolved Reactive Phosphorus) concentrations below 0.019–0.033 mg/L, depending on the river type. In general, high order, low slope rivers and streams required more stringent nutrient criteria than steep, low order streams. The link between nutrient concentrations and biological indices were weak and likely suffer from the loss of information from summarising an entire assemblage into a single numeric. We consider that the derivation of nutrient criteria for waterways should also examine the individual relationships with the taxa in a river system to establish protection for a desired proportion of taxa.