The present study took a meta-analytic approach to investigate whether students' engagement acts as a mediator in the association between affective teacher-student relationships and students' achievement. Furthermore, we examined whether results differed for primary and secondary school and whether similar results were found in a longitudinal subsample. Our sample consisted of 189 studies (249,198 students in total) that included students from preschool to high school. A distinction was made between positive relationship aspects (e.g., closeness) and negative relationship aspects (e.g., conflict). Meta-analytic structural equation modeling showed that, overall, the associations between both positive relationships and achievement and negative relationships and achievement were partially mediated by student engagement. Subsequent analyses revealed that mediation is applicable to both primary and secondary school. Only the direct association between positive relationships and engagement was stronger in secondary school than in primary school. Finally, partial mediation was also found in the longitudinal subsample.