The closeness of electoral races has consistently been argued to increase voter turnout. Several studies have confirmed this finding by using double-ballot contests as a measure of ‘actual’ closeness between candidates or parties; yet, research has so far largely been limited to legislative elections and single country-studies. In this paper, we analyse the effects of electoral closeness on turnout differences in presidential elections and provide one of the first cross-national studies of this kind. We develop a modified model of turnout in presidential elections based on voter awareness and learning and test its propositions on a new data set covering all popular presidential elections in 16 European democracies, 1965-2020. Overall, we find that the first-round margin of closeness between leading candidates is associated with a significant increase in turnout; however, incumbents as frontrunners, presidential powers, and non-concurrent legislative elections prove to be more important predictors. Thereby, the results are stable for different operationalisations of the dependent and independent variables. We furthermore compare our findings with estimations based on pre-electoral polls; surprisingly, actual closeness between frontrunners in the first round performs better and inclusion of polling data into a combined model only marginally improves model fit. Our results offer new perspectives on voter turnout in semi-presidential democracies and contribute to scholarship on the political and institutional determinants of electoral behaviour.