Curiosity theories converge in explaining curiosity as a function of people’s knowledge gaps. However, they diverge in their predictions about the magnitude of these gaps that elicits peak curiosity. The novelty theory posits that curiosity is highest when knowledge gaps are large. The information-gap theory posits that curiosity peaks when knowledge gaps are moderate. Finally, the region of proximal learning theory posits that curiosity peaks when knowledge is almost complete, with just a small gap remaining. Here, we experimentally tested these theories by illustrating knowledge gaps to participants and measuring their subsequent information-seeking behavior. Across three experiments (n = 200 each), participants first completed a pretest assessing their prior knowledge about independent chapters on a topic. They were then randomly assigned to one of two conditions: an experimental group, in which knowledge gaps were explicitly illustrated, or a control group, in which they were not. Participants then decided whether to read or skip each chapter, for which their knowledge gaps were either illustrated or not. We found increased information-seeking behavior when illustrating large-to-moderate knowledge gaps compared to the control group. We found no evidence that illustrating small knowledge gaps increased information-seeking behavior. Critically, these effects were robust across time. Our findings provide converging evidence that making moderate-to-large knowledge gaps explicit elicits curiosity, reflected in higher information-seeking behavior. Together, these results support the novelty and information-gap theories, suggesting that curiosity arises when people detect a large-to-moderate knowledge gap.