Adaptive tutoring trial offers evidence, not a verdict on chatbots

Read on ⮛

Research announcement: 24 June 2026 · England · Reviewed: 12 September 2026

An Education Endowment Foundation evaluation of Maths-Whizz involved 63 primary schools in England, with pupils in Years 2–5. The independently evaluated randomised trial found the equivalent of one additional month’s maths progress on average, and two months for pupils eligible for free school meals.

Maths-Whizz adapts activities and feedback to the learner. As the EEF makes clear, it is not a generative AI tutor.

For educators

The distinction matters when considering claims about “AI in education”. This finding concerns a particular programme, subject and age group; it does not establish that open-ended chatbots improve learning. When evaluating a tool, ask what was tested, with whom, against which comparison and for which outcome.

Read the EEF findings and evaluation links.

Back to AI in Schools updates

Explore the connections

Browse the Insight Library · Listen to Fact & Friction

We use cookies to improve your experience. By clicking ‘Accept’, you agree to our use of cookies. Privacy Policy