We analyse the role of changes in the geographic concentration of immigrants in shaping teachers’ assessment of students’ performance. By using data on Italian students attending the fifth grade, we adopt an IV estimation strategy and, by controlling for student performance in blindly scored tests of proficiency, find that an increasing presence of immigrants in the local population negatively affects the way teachers evaluate immigrant students, as opposed to their peers, in non-blindly graded tests. We also reveal that the effect is mainly driven by schools located in smaller communities and in areas with lower overall levels of educational attainment and that it is unlikely to be related to the conduct of immigrant students who live in the areas experiencing sizable increases in immigration flows. In addition, older teachers tend to be more biased.