Researchers Publish Review Finding AI Can Help or Hinder Student Thinking

Generative AI has been sold to schools as both a study aid and a threat to independent learning. A new systematic review of 173 international studies finds that neither label is enough: AI can support critical thinking and problem-solving when students use it to question, compare, and build their own reasoning, but it can weaken those skills when it becomes a source of ready-made answers.

The review, published in the Journal of Information Technology Education: Research, examined research on generative AI in schoolwork and education published from 2022 through spring 2025. Its central conclusion is not that the technology has one fixed educational effect. The interaction matters: prompting students to interrogate an answer or weigh competing views is different from letting a chatbot complete the intellectual work.

The results show why a single verdict is hard to defend. Of 139 studies that considered critical thinking, 67 found only positive effects from AI use, 33 found only negative effects, and 39 found both. The balance was also mixed for problem-solving: 57 of 103 studies reported exclusively positive effects, while 23 found exclusively negative effects and another 23 reported both.

That conclusion comes with major limits. Eighty percent of the reviewed studies involved university or college participants, leaving much less evidence about children and younger teenagers. The gap matters because younger students are still developing how they process, assess, and reason through information; findings from adult learners may not transfer cleanly.

Three limits on the headline result

• Most of the underlying studies used self-reported data rather than objective measurements tracking changes in learning or cognitive skills over time.
• The research was geographically concentrated in East, Southeast, and South Asia, while Africa, South America, and Central Asia were seriously underrepresented.
• The review covers a rapidly changing period, from 2022 through spring 2025, rather than a long-term record of how classroom habits develop with AI.

For teachers and school leaders, the review shifts the question from whether students should use AI at all to what role the tool plays inside an assignment. Work that asks students to test an AI response, identify its assumptions, or compare it with other evidence may exercise judgment. Work that accepts a generated response as the endpoint may conceal whether judgment happened at all.

The review does not settle which classroom rules work best, nor does it establish the long-term effects of AI use on younger pupils. It does establish a narrower warning: treating generative AI as automatically educational—or automatically corrosive—misses the part educators can still shape, which is how much thinking the student must retain.