In three experiments, we used attention capture effects (of cues) to investigate the emergence of human’s top-down attentional control settings after learning versus verbally instructing target-defining features during visual search. In Experiment 1, participants’ learning of a target-defining color prompted attentional control settings indistinguishable from those prompted by verbal instructions – as indicated by their effects on attention capture. In Experiment 2, color and orientation both independently defined the target. In line with prior learning experiments, participants reported that they learned and used color to search for the target. Consequently, only color cues matching target search settings captured attention, whereas cues with the same orientation as the target did not. To further investigate the dependence of attentional control settings on learned target features, in Experiment 3, participants had to relearn orientation as the new target-defining feature after they had previously learned to search for a color target. The cues’ attention effect suggested that by informing participants that color was no longer the target’s defining feature, participants almost immediately deactivated their previously effective attentional control setting for target color, even before they learned the new target feature. After learning the new target feature (orientation), only cues matching this search setting captured attention, with no capture by cues of the previous targets’ color. These results indicate that attentional control settings react fast and flexibly to learned target features and highly resemble those based on verbal instruction.