US Education Department Guidance Requires AI Edtech to Be Judged on Learning Outcomes, Not Usage

The US Department of Education issued guidance last week separating recreational technology from instructional technology and raising the bar for instructional technology, including AI. It calls for judging products by demonstrated learning outcomes rather than usage, building evidence into renewals, and expecting vendors to publish independent evaluations and disclose what their products cannot do. The article also cites a two-year randomized trial across 18 Tennessee middle schools: 96% of students tried an AI tutor at least once, but turned to it in only 17% of the moments when they had gotten something wrong.

The guidance shifts proof of value for AI education products from usage to learning outcomes, directly affecting school purchasing and renewal conditions and the evidence vendors must supply. For teachers and schools, the hard part is implementation: the Tennessee trial cited in the article shows the tool was one click away yet students used it in only 17% of the moments when they had gotten something wrong, indicating the gap between a capable tool and a learning outcome is closed by design and adult decisions. The article also notes the evidence is early and does not provide the guidance's full provisions or enforcement details, so its practical force remains to be seen. School purchasing and renewals will weigh demonstrated learning outcomes and independent evaluations more than activation or usage. AI education vendors must publish independent evaluations and disclose product limitations, raising the cost of evidence.