Two RAND surveys found AI use split sharply by school poverty

The classroom note

Two RAND Corporation surveys, run through its American Educator Panels, asked teachers and principals directly whether they were using AI tools, rather than asking families or students to guess. The first, published 17 April 2024, surveyed a nationally representative 1,020 teachers and 231 districts in fall 2023 and interviewed 11 district leaders that winter. It found 18% of K-12 teachers reported using AI for teaching, with another 15% having tried it at least once, concentrated among middle and high school teachers and those teaching English language arts or social studies. A follow-up report published 11 February 2025, covering the same 2023-2024 school year, put overall teacher use at 25% and added principals to the picture for the first time: nearly 60% reported using AI for their own work.

What the evidence says

Both reports are teacher and principal self-report, not observation of classroom practice, and both were funded in whole or part by the Gates Foundation, with the second report adding the Charles and Lynn Schusterman Family Foundation and the Walton Family Foundation as co-funders, organisations that also support AI-in-education initiatives elsewhere. The second report's subject breakdown found English and science teachers roughly twice as likely to report AI use as mathematics or general elementary teachers, at around 40% against 20%. Both surveys converge on a poverty gap: the earlier report found urban districts least likely to have delivered AI training by the end of the school year, and the later one found teachers and principals in higher-poverty schools less likely to report using AI, with only 13% of higher-poverty-school principals reporting their school gave staff guidance on AI use, against 25% in lower-poverty schools.

The implementation question

Guidance here is a specific, checkable thing: a policy document, training session or written expectation a principal can point to, not a general awareness that AI exists. The gap the surveys report is not simply about who wants to use AI tools; it tracks which schools have the staff time, technology budget and administrative capacity to write and deliver that guidance in the first place. A district with fewer resources is being asked to make the same policy decisions as a wealthier one, with less support to do it.

What holds and what fails

The finding holds as two independently fielded, methodologically transparent surveys of a common national panel, converging on the same direction of gap across two school years. It does not establish why the gap exists, whether it is closing, or what effect it has on pupils; self-reported use also cannot distinguish a teacher who tried a chatbot once from one using it daily, since neither report surveys students directly.

• Is this figure teacher-reported use, or independently observed practice?
• Does our school's poverty profile match the schools where this survey found less guidance?
• Who funded this survey, and does that shape which questions were asked?

A repeated, transparent survey is a genuine early-warning system for unequal adoption; it is not, on its own, a measure of what any of this use is doing for pupils.