Students can generate teaching materials for others. However, solid evidence of the learning effect for the student creating the material is needed. This meta-analysis aims to explore current evidence of the learning-by-teaching potential effect on students' learning of the content they teach. A total of 23 articles were included in the meta-analysis, providing 62 comparisons be-tween an intervention in which students were required to create a teaching material and an alternative condition. Possible moderators were examined as well (i.e., product, educational level, content area, and access to source material). Results show statistically significant differences in favour of creating teaching materials when compared to business-as-usual or alternative in-terventions, with an overall effect of 0.17. However, the type of product and the access to source material significantly moderate the findings. Audio-visual and visual materials considerably outperform textual materials. Having no access to the source material is better than having full access to it. No publication bias was detected. However, the type of control group moderates the findings: creating teaching materials shows a significant effect when compared to nonbeneficial interventions rather than to other expected beneficial interventions. Nonbeneficial interventions refer to business-as-usual or alternative interventions that are not expected to influence the outcome much, while beneficial interventions are those that are known or expected to have a positive effect. The findings are discussed based on the different views on the underlying learning -by-teaching mechanisms. Limitations as well as implications for practice, policy, and future research are underlined.