We analyse gender differences in children’s aspirations and academic beliefs that affect their academic potential. We use administrative data on the test scores of a primary school in rural China and a novel survey instrument in the evaluation of a randomised controlled trial of a role-model intervention. Our intervention increases Chinese test scores by 0.116 standard deviations, which is driven by boys’ 0.33–0.35 standard-deviation performance boost that persists into 1 year after the treatment. Girls only experience a short-term increase in performance within a week after the treatment. The impact of the treatment is especially pronounced for children starting with low aspiration, self-efficacy, or cognitive ability. While there is no average impact on Mathematics test scores, we find that boys holding traditional gender attitudes and having a high self-efficacy, and girls of high self-efficacy and aspirations, perform better in response to the treatment. Aspirations, confidence, and gender attitudes account for parts of our findings.