In 2007, supplementation with the trace element selenium in a trial was unexpectedly found to be associated with an excess risk of type 2 diabetes. Given the concerns raised by these findings and the large number of recent studies on this topic, we reviewed the available literature with respect to this possible association. In this paper, we assessed the results of both experimental and nonexperimental epidemiologic studies linking selenium with type 2 diabetes incidence. Through a systematic literature search, we retrieved 50 potentially eligible nonexperimental studies and 5 randomized controlled trials published through June 11, 2018. To elucidate the possible dose–response relation, we selected for further analysis those studies that included multiple exposure levels and serum or plasma levels. We computed a pooled summary risk ratio (RR) of diabetes according to selenium exposure in these studies. We also computed a RR for diabetes incidence following supplementation with 200 µg/day of selenium compared with placebo in trials. In the nonexperimental studies, we found a direct relation between selenium exposure and risk of diabetes, with a clear and roughly linear trend in subjects with higher plasma or serum selenium levels, with RR at 140 µg/L of selenium exposure compared with a referent category of < 45 µg/L equal to 3.6 [95% confidence interval (CI) 1.4–9.4]. A dose–response meta-analysis focusing on studies with direct assessment of dietary selenium intake showed a similar trend. In experimental studies, selenium supplementation increased the risk of diabetes by 11% (RR 1.11, 95% CI 1.01–1.22) compared with the placebo-allocated participants, with a higher RR in women than in men. Overall, results from both nonexperimental and experimental studies indicate that selenium may increase the risk of type 2 diabetes across a wide range of exposure levels. The relative increase in risk is small but of possible public health importance because of the high incidence of diabetes and the ubiquity of selenium exposure.