Species distribution models (SDMs) are widely used to predict range shifts but could be unreliable under climate change scenarios because they do not account for evolution. The thermal physiology of a species is a key determinant of range and thus incorporating thermal trait evolution into SDMs might be expected to alter projected ranges. We identified a genetic basis for physiological and behavioural traits that evolve in response to temperature change in natural populations of threespine stickleback (Gasterosteus aculeatus). Using these data, we created geographic range projections using a mechanistic niche envelope approach under two climate change scenarios. Under both scenarios, trait data was either static (‘no evolution’ model), allowed to evolve at observed evolutionary rates for the trait (‘evolution’ model), or allowed to evolve at a rate of evolution scaled in association with the variance that is explained by quantitative trait loci (QTL; ‘scaled evolution’ models). We show that incorporating these traits and their evolution substantially altered the projected ranges for a widespread panmictic marine population, with over 7-fold increases in area under climate change projections when traits are allowed to evolve. Evolution-informed SDMs should improve the precision of forecasting range dynamics under climate change, and aid in their application to management and the protection of biodiversity.