Metabolism is a key property of lake ecosystem functioning, but logistical challenges make it difficult to estimate across remote regions. The steady-state dissolved oxygen (DO) stable isotope method (18O method) estimates metabolism from discrete water samples and thus enables large-scale surveys. However, this method relies on the assumptions that the upper mixed layer DO saturation (DO%) relative to its isotopic composition (δ18ODO) is at a steady state and that an increase in DO% results in a proportional decrease in δ18ODO. The applicability of these assumptions has not been broadly assessed for small, low-productivity lakes with predominantly benthic metabolism. We evaluated the 18O method in these types of systems by surveying 184 Arctic lakes in Sweden and found that the method consistently produces realistic estimates of metabolism in well-mixed conditions and when water temperatures were relatively stable. Under such conditions, results from the 18O method agreed with those from the free-water diel DO method, and rates derived from both methods responded similarly to environmental drivers. In contrast, we found that the 18O method frequently generated unrealistic metabolic rates when temperatures were rising. Increasing temperatures may increase DO% irrespective of δ18ODO in the upper mixed layer and promote lake stratification, both violating the assumptions of the 18O method and preventing benthic metabolism from being integrated by surface water samples. We conclude that the 18O method is a powerful tool for studying metabolism in Arctic lakes across large spatial gradients, provided that temperature dynamics and vertical stratification are considered.