People can learn to pay attention to locations and simple features frequently associated with visual search targets. Here, we tested how such statistical regularities influence attentional selection during real-world object search. Results showed that participants acquired long-lasting attentional biases for a real-world object that was frequently their target (Exp. 1, N=44). This attentional bias was specific to that particular object, with no benefits for a novel object from the same category (Exp. 2, N=32). In contrast, learning to prioritize at least two exemplars from one category rapidly transferred to novel objects from the same category (Exp. 3, N=72). This indicates that people can tune their attention broadly, to the level of object categories, when recent experience suggests it could be useful. These results demonstrate that the breadth of attentional tuning during real-world object search can be flexibly adjusted based on recent experience to most effectively support current task demands.