Review of 12 studies shows AI tools boosted elementary pupils' learning, though the effect may be inflated

The authors call that effect size particularly large and suggest it may be explained by the small sample sizes of the included studies and potential publication bias. The figure should therefore be read as an upper-end estimate rather than a settled benefit.

Subgroup results differed: AI had less effect in Mathematics, while chatbots and automated formative assessment systems were the most effective types. The authors say long-term effects on learning still need study.

The review set out to synthesize the available evidence on how AI affects elementary school students' learning. It restricted itself to experimental and quasi-experimental designs from 2020 to 2024, and 12 of 272 studies met the inclusion criteria.

The headline result is a pooled Hedges' g of 0.91 with a confidence interval of 0.55 to 1.27. The authors say this indicates AI integration has the potential to support learning, but they treat the size cautiously because the included studies had small samples and publication bias is possible.

A moderator analysis found variation by subject and tool type. AI had less effect in Mathematics, while chatbots and automated formative assessment systems were the most effective. The authors recommend future studies examine long-term impact.