Over the last decade reports of animal poisoning following accidental consumption of neurotoxin-producing benthic cyanobacteria (mainly Phormidium spp.) have increased. There is a need for rapid and cost-effective tools to survey benthic cyanobacteria. In this study we assessed the performance of the BenthoTorch, a fluorometric probe that provides in situ estimations of cyanobacteria, diatoms and green algae biomass in biofilms. Biofilms = 288) were analysed from two rivers in France and eight in New Zealand. Correlations between chlorophyll-a measured using the BenthoTorch and spectrophotometry were higher for thin (<2 mm) compared to thick (>2 mm) biofilms (r(2) = 0.58 and 0.27 respectively; p < 0.001). When cyanobacteria represented less than 50% of the total biomass (based on biovolumes), microscopic and BenthoTorch compositional estimations were significantly correlated (r(2) = 0.53, p < 0.001). Conversely, there was no correlation when cyanobacteria exceeded 50% of the total biomass. Under this scenario diatoms were overestimated. Our results suggest that the observed biases occur because the BenthoTorch only measures the upper biofilm layer and it underestimates the biomass of phycoerythrin-containing cyanobacteria. To improve the performance of this sensor and render it a useful tool for a rapid evaluation of benthic cyanobacterial biomass in rivers, we propose that: (i) the algorithms based on the LEDs responses currently available on this tool need revision, (ii) new excitation wavelengths should be included that allow the fingerprints of phycoerythrin-containing cyanobacteria to be discriminated, and (iii) a sensor that penetrates the biofilms is needed to obtain more accurate estimates of cyanobacterial biomass. (C) 2016 Elsevier Ltd. All rights reserved.