In three experiments (Ns = 327/137/210), we investigated whether test grades and elaborated feedback in a stereotypically male (Math) and a stereotypically female subject (German) are biased by the student's gender. For this purpose, pre-service teachers graded and provided written feedback on tests which were allegedly from boys or girls. In addition, participants' belief in stereotypes was measured in Study 1 and 2 and manipulated in Study 3 to test its moderating role. A meta-analysis across the three studies confirmed the following pattern: a small to moderate stereotype-contrasting grading bias, if the evaluators endorsed stereotypes, but no bias if they did not. Tests from the gender that, according to the stereotype, is weaker in the domain, were graded better. Study 1 and 3 further showed that the supposedly weaker gender received more elaborated feedback. The results are discussed in terms of shifting standards and previous findings in gender bias in school.