Effective societal responses to rapid climate change in the Arctic rely on an accurate representation of region-specific ecosystem properties and processes. However, this is limited by the scarcity and patchy distribution of field measurements. Here, we use a comprehensive, geo-referenced database of primary field measurements in 1,840 published studies across the Arctic to identify statistically significant spatial biases in field sampling and study citation across this globally important region. We find that 31% of all study citations are derived from sites located within 50 km of just two research sites: Toolik Lake in the USA and Abisko in Sweden. Furthermore, relatively colder, more rapidly warming and sparsely vegetated sites are under-sampled and under-recognized in terms of citations, particularly among microbiology-related studies. The poorly sampled and cited areas, mainly in the Canadian high-Arctic archipelago and the Arctic coastline of Russia, constitute a large fraction of the Arctic ice-free land area. Our results suggest that the current pattern of sampling and citation may bias the scientific consensuses that underpin attempts to accurately predict and effectively mitigate climate change in the region. Further work is required to increase both the quality and quantity of sampling, and incorporate existing literature from poorly cited areas to generate a more representative picture of Arctic climate change and its environmental impacts.