This classroom-based study employs a mixed-methods approach to exploring both short-term and long-term effects of Criterion feedback on ESL students' development of grammatical accuracy. The results of multilevel growth modeling indicate that Criterion feedback helps students in both intermediate-high and advanced-low levels reduce errors in eight out of nine categories from first drafts to final drafts within the same papers (short-term effects). However, there is only one error reduction of statistical significance in the category of Run-on Sentence from the first drafts of the first paper to the first drafts in the subsequent papers for both levels of students (long-term effects). The findings from interviews with the participants reveal students' perceptions of Criterion feedback and help us understand the feedback effect. Implications for a more effective use of AWE tools in ESL classrooms are discussed.