Teachers' perceptions of their students' academic skills can affect students' achievement and may be influenced by unrelated student characteristics such as socioeconomic status (SES). In this ad hoc randomized controlled trial, teachers (n = 28) were randomly assigned to receive training on using assessment to guide literacy instruction, Assessment-to-Instruction (A2i), or on Math PALS (control). Teachers rated students' (n=446) academic competence. A2i teachers' ratings did not vary by SES, and their ratings correlated more strongly with students' literacy and mathematics assessment scores compared with those of the control teachers. Control teachers generally underestimated lower SES students' academic competence; underestimation was greater at more affluent schools. Teachers' ratings of students' academic competence predicted reading and mathematics outcomes. Thoughtful use of assessments to guide instruction appeared to improve the precision of teachers' ratings of students' academic competence, improve student outcomes, and reduce potential teacher biases about children from higher-poverty families.