Voting Advice Applications (VAAs) fulfill different needs for different citizens. In national elections, the majority of users can be characterized as politically sophisticated citizens who use VAAs for entertainment purposes and confirmation of their party preference, but a significant minority uses VAAs to learn about politics and make an informed vote choice. VAAs might, however, play a different role in second-order elections, since in these elections campaign dynamics and information supply are very different. In the current research, we applied latent class analysis on user data from a widely used Dutch VAA (Kieskompas) for a supranational and several subnational elections in the Netherlands, to test if an extant typology of VAA users for national elections could be replicated. We find that the typology can be replicated for most of these elections, but also that the relative size of the