In response to a widespread belief within the legal system that cross‐examination is instrumental in uncovering the truth, we examined the effect of cross‐examination questioning on the reports of children who had—and had not—been coached to lie. A group of children, aged 6–11 years (N = 65), played three computer games with one of their parents. For half of the pairs, the parents—who acted as confederates—coached their children to make lies of commission concerning the occurrence of two target activities. For the remaining pairs, these two target activities actually occurred, and there was no coaching. Immediately afterwards, children were interviewed about the two activities. Those who—correctly or incorrectly—reported that both activities occurred were retained for the final sample (n = 56); these children were then interviewed again with both neutral questions and cross‐examination‐style challenges. Neither style of questioning elicited responses that discriminated between liars and truth‐tellers: although the accuracy of children who were lying increased in response to cross‐examination questions, the accuracy of truth‐telling children saw a corresponding decrease. When asked neutral questions, children's responses tended to be consistent with their earlier responses, whether or not those responses were lies. These findings raise important questions about the function that cross‐examination might serve in trials involving child witnesses. Copyright © 2016 John Wiley & Sons, Ltd.