Access to an AI tutor does not establish that a learner will use it productively. An August 2026 working paper reports a two-year randomized study across 18 Tennessee middle schools. Students assigned to Khan Academy with Khanmigo made mathematics gains, but substantive engagement with the AI tutor was limited. The authors explicitly state that the design cannot isolate Khanmigo’s additional contribution from the wider intervention. This is preliminary school-based evidence, not a comparison of private AI mentors. (edworkingpapers.com)
Our inference for buyers: Interview mentors about how they respond when a child avoids a difficult step, repeatedly asks AI for finished answers or loses interest. A convincing answer should describe teaching actions and follow-up—not merely access to better software.
Sources cited on this page
- One Click Away: AI Tutoring with Khanmigo in a Two-Year School Experiment ↗edworkingpapers.com