Recent research indicates that perceptual learning (PL) interventions in real-world domains (i.e., mathematics, science) can produce strong learning gains, transfer, and fluency. Although results on domain-relevant assessments suggest characteristic PL effects, seldom have real-world PL interventions been explicitly tested for their effects on basic information extraction. We trained participants to classify Chinese characters, based on either (1) overall configurations (structures), (2) featural relations (components), or (3) non-relational information (stroke-count control). Before and after training, we tested for changes in information extraction using a visual search task. Search displays contained all novel exemplars, involved manipulations of target-distractor similarity using structures and components, and included heterogeneous and homogeneous distractors. We found robust improvements in visual search for structure and component PL training relative to the control. High-level PL interventions produce changes in basic information extraction, and sensitivity induced by PL for both relational structure and specific components transfers to novel structural categories.