To reduce sensory uncertainty, humans combine cues from multiple senses. However, in everyday life, many co‐occurring cues are irrelevant to the task at hand. How do humans know which cues to ignore? And does this ability change with development? This study shows the ability to ignore cross‐modal irrelevant information develops late in childhood. Participants performed a sound discrimination task, with or without an irrelevant visual flash, presented synchronously in front of them. Adults ignored the irrelevant visual information, while 7‐ to 10‐year‐olds' responses were biased toward the flash location. The findings show that acquiring mature cue combination mechanisms is a multifaceted process that includes learning to ignore irrelevant cues, as well as to optimally combine relevant cues.