Value-modulated attentional capture (VMAC) refers to a process by which a priori neutral stimuli gain attentional priority when associated with reward, independently of goal or stimulus-driven attentional control. Although VMAC is considered an automatic and implicit process, the role of awareness of the stimulus-reward contingency on its learning process remains unclear at best. In a well-powered replication of a previous study, we found that VMAC is absent when participants are not explicitly informed about the stimulus-reward contingency in the pre-task instructions. In a second experiment, we show that when instructions are manipulated between groups, only the instructed group shows VMAC. Interestingly, although the no-instruction group did not show VMAC at the group level, participants who became aware of the stimulus-reward contingencies did nevertheless show robust VMAC at the end of the task. Meta-analytic evidence further supports our conclusion by showing that studies that include instructions about the stimulus-reward contingencies yield significantly larger VMAC effects. Taken collectively, these findings suggest that the learning process behind VMAC may not be entirely implicit.