173 studies on AI in classrooms — and 80% tested university students
Norway just published the clearest look yet at what generative AI does to how children think. The honest answer is that nobody has run the study that would tell us.
A review of 173 international studies finds a clean dual effect from AI in the classroom — and a hole where the evidence on children should be.
The review, led by Hilde Storrøsæter at NTNU's Department of Teacher Education with colleagues from Czech universities, analyzed research published from 2022 through the spring of 2025 on generative AI in schoolwork. Her conclusion is unusually even-handed for a field this polarized: "AI in and of itself does not make us any better or worse at thinking," she said. Used to challenge a claim, compare viewpoints or build a line of reasoning, AI supports exactly the skills teachers say they want. Used as a shortcut to a ready-made answer, it lines up with overdependence, more superficial work and weaker independent problem-solving. The counts back the ambiguity rather than resolving it: of 139 studies on critical thinking, 67 reported only positive effects, 33 only negative, and 39 both. Problem-solving split 57 positive, 23 negative, and 23 mixed. The literature mostly agrees that the answer is "it depends," and mostly declines to say on what.
What the review cannot say is what happens to a ten-year-old. In 80% of the studies it examined, the participants were university or college students. "We can't simply assume that what applies to a 22-year-old university student also applies to a 10- or 14-year-old," Storrøsæter said — and if AI takes over the thinking process before pupils have built their own critical-thinking and problem-solving skills, the effects could be entirely different. Most of the reviewed studies also rest on self-reporting rather than measured change over time, the weakest evidence available. The research base is geographically lopsided too: 28 studies from South, Southeast and East Asia (primarily China), nine each from Spain and the UK, 15 from the United States, and almost nothing from Africa, South America or Central Asia. Our confidence about the most consequential classroom technology in a generation rests largely on surveys of adults in a handful of rich countries.
The takeaway is not that schools should ban AI. It is that the policy being written right now is running ahead of the evidence, and the gap is specific, measurable and unfunded. We have a 173-study review telling us the missing study is a cohort of schoolchildren followed over time, and no institution has commissioned the follow-up. Storrøsæter's prescription is narrower and harder than a ban: teacher education has to carry the load, because AI literacy is not a tool you hand a child, it is a judgment you teach them.
Norwegian 9- to 18-year-olds are already leaving Google for AI, and the data on what that does to them does not exist.
The Norwegian Media Authority's 2026 media-habits survey puts Google search use among 9- to 18-year-olds at 47%, down from 72% in 2024. The NTNU team's own figure is that 39% of Norwegian 11- and 12-year-olds use AI, and the youth information service ung.no lost a quarter of its traffic — 24 million visits in 2024 falling to 18 million in 2025 — as teenagers started asking chatbots instead of reading the state's answers. The shift is not being shaped by any curriculum; students are absorbing it faster than schools can write rules for it. Uncapped, that produces the split Storrøsæter warns about: children who use AI critically and as a tool, and children who use it to finish faster, with the gap widening through every year of school.
The classroom evidence had already been pointing this way — we covered the OECD finding that students leaning on AI score a year and a half behind and MIT's warning on cognitive surrender in AI-era classrooms. What is new here is the size of the blind spot: 173 studies, and still no cohort of children followed over time.
What to watch: whether an education ministry funds the missing cohort study, or keeps legislating from research on undergraduates.
If your school introduced AI rules this year, were they written with teachers or handed to them? Tell us in the comments.
Sources: Norwegian SciTech News (NTNU) · Phys.org · Dual Nature of GenAI: systematic review, Informing Science · Norwegian Media Authority — Barn og medier 2026
