A two-year cluster-randomized trial in 18 Tennessee middle schools tested Khan Academy’s Khanmigo AI tutor during existing daily remedial mathematics sessions. The system was configured to coach students rather than provide answers. Students assigned to use Khanmigo gained 1.3 national percentile ranks in math per term, equivalent to roughly 0.06 to 0.08 standard deviations over a school year; the estimated effect for a full year of active participation was 0.14 standard deviations. These gains were similar to those associated with Khan Academy practice without AI assistance. Access did not translate into sustained use: 96 percent of students tried Khanmigo at least once, but the median student messaged it on only one-third of practice days and in just 17 percent of sessions where they made a mistake. When students did send messages, they were usually bare answers or responses to suggested prompts rather than substantive mathematical discussions. The authors identify engagement as the main constraint, arguing that effective AI tutoring requires students to use the tutor meaningfully, not merely have access to it.
