Buying AI for Education: A Standard for Evidence, Privacy and Cost

Universities and school systems are signing multiyear AI contracts with little independent evidence, uneven privacy and accessibility terms, and prices that can hide much of the cost. This white paper reviews what is known about edtech evidence, vendor data practices, usage and pricing, and proposes a Five-Gate Purchase Standard, covering purpose, proof, protection, price and exit, that leaders can apply to any AI purchase or renewal.

Key findings

1. A Stanford review found only 20 high-quality causal studies among more than 800 papers on AI in schooling, and none on student use of AI in US K-12 classrooms.
2. Among 150 widely used classroom technologies, 40 percent of tools built for education and 2 percent of consumer tools had evidence aligned with US federal tiers, and only 2 percent reached the top level.
3. In 32 percent of studies in an analysis of a US federal evidence database, a significant positive effect on a developer or researcher measure appeared beside null effects on every independent measure.
4. Only 41 percent of 607 surveyed US districts require vendors to document evidence-based design, and 11 percent formally require WCAG 2.1 AA accessibility in procurement policy.

The paper has 2 more findings, the full evidence review and the framework.

A chart from the paper

What is inside

1. Executive summary and key findings
2. Why AI buying became a board-level question
3. What the evidence shows
4. Where the evidence is weak or contested
5. A practical framework: the Five-Gate Purchase Standard
6. Case snapshots
7. What to measure
8. Risks and open questions
9. Recommendations for universities, educators, edtech leaders and policymakers
10. Methodology, limits and 34 references

One recommendation for each audience

This paper is desk research completed in October 2026. We searched for peer-reviewed studies, working papers, official and regulator documents, sector surveys and major news reporting on edtech evidence, vendor data practices, accessibility, pricing and contracts, and used only sources we opened and read. We preferred primary and independent sources and noted where a source has a commercial interest. Figures marked as our calculation are derived from cited numbers. The paper is not original fieldwork, a legal opinion or a product review, and the Five-Gate Purchase Standard is our synthesis, not a tested instrument. The cover picture is an AI-generated illustration. Send a correction · Research Standards