People can learn, and utilize, not only static but also dynamic (cross-trial) regularities in the positioning oftargetitems in parallel, ‘pop-out’ visual search. However, while static target-location learning also works in serial search, acquiring dynamic regularities seems prevented by the demands imposed by item-by-item scanning. Also, questions have been raised regarding a role of explicit awareness for utilizing (at least) dynamic regularities to optimize performance. The present study re-investigated whether dynamic regularities may be learned in serial search when regular shifts of the target location occur frequently, and whether such learning would correlate with awareness of the dynamic rule. To this end, we adopted the same regularity used by Yu et al. (2023) to demonstrate dynamic learning in parallel search: a cross-trial shift of the target location in, e.g., clockwise direction within a circular array in 80% of the trials, which was compared to irregular shifts in the opposite (e.g., counterclockwise; 10%) or some other, random direction (10%). The results showed that ⅔ of participants learned the dynamic regularity, with their performance gains correlating with awareness: the more accurately they estimated how likely the target shifted in the frequent direction, the greater their gains. Importantly, part of the gains accrued already early during search: a large proportion of the very first and short-latency eye movements was directed to the predicted location, whether or not the target appeared there. We discuss whether this rule-driven behavior is causally mediated by conscious control. (248 words).