Walk into almost any staffroom this year and you’ll find the same argument happening in different words. One teacher will tell you a student handed in an essay that was clearly written by a chatbot, not a fifteen-year-old. Another will tell you that same chatbot spent twenty minutes patiently re-explaining quadratic equations to a kid who’d given up on the subject months ago. Both teachers are right. That’s the uncomfortable part.

Artificial intelligence has moved from novelty to fixture in classrooms faster than almost any education technology before it — solving equations instantly, drafting essays in seconds, answering questions a tired student might otherwise never ask out loud. The tools are genuinely useful. The question nobody’s fully answered yet is what they’re quietly doing to the students using them.

The Numbers Say Everyone’s Worried — Including the Students
This isn’t a debate running on anecdotes anymore. A national NPR/Ipsos poll of American K-12 teachers, published in June 2026, found that more than half — 55% — see AI mainly as a shortcut students use to avoid doing real work, and 54% say it’s actively making it harder for students to build critical thinking skills. Three in four teachers believe AI’s impact on education will be bigger than any past wave of classroom technology. And yet only about a third say their school actually has formal guidance on how students are allowed to use it — which means most teachers are watching this unfold with no rulebook to fall back on.

Higher education isn’t faring any better. A January 2026 survey by the American Association of Colleges and Universities found 95% of college faculty worried about students becoming overly reliant on AI, with 73% saying they’d personally dealt with an AI-related academic integrity case, and 83% predicting it will shorten students’ attention spans over time.

What’s striking is that students aren’t oblivious to any of this — many share the exact same worry about themselves. RAND Corporation’s American Youth Panel, surveying more than a thousand students aged 12 to 29, found that AI use for homework climbed from 48% to 62% of students between May and December of 2025 alone, driven mostly by middle and high schoolers. In that same window, the share of students who believed AI was hurting their own critical thinking rose from 54% to 67%. As Heather Schwartz, who leads education research at RAND, put it: students are clearly embracing AI as a learning tool, but they’re just as clearly conflicted about what that means for their own learning. They’re using it to look things up, get explanations, brainstorm essays — and increasingly suspecting, in the same breath, that it’s costing them something.

Is Using It the Same as Cheating?
For a certain kind of teacher, the answer is close to yes. The concern isn’t really about the tool — it’s about what gets skipped when a student submits AI-generated work without wrestling with the underlying concept first. Learning was never really about arriving at the right answer; it was about the mental effort of getting there, the dead ends, the half-right guesses that eventually sharpen into understanding. Hand that process to a machine, and what’s left to hand in is a finished answer with none of the thinking that’s supposed to come attached to it.

Teachers in the UK’s National Education Union have voiced almost identical frustration. One described watching AI produce, in their words, “sub-standard slop” that students submit anyway, simply because they haven’t been trained to use the tool properly in the first place. Another flagged something more basic and more alarming — that AI will confidently invent an answer when it doesn’t actually know one, and a student without the background to spot that has no way of catching the error.

But a Lot of Teachers Are Using It Too — and Liking It
Here’s the part that complicates any simple “AI is ruining school” narrative: most teachers are using AI themselves, and largely say it’s made them more productive. The NPR/Ipsos poll found teachers’ own top uses were building classroom materials (69%) and writing or planning lessons (52%) — administrative and preparation work, more than direct instruction, but real time saved all the same.

And when researchers looked closely at how AI actually behaves inside a well-designed lesson, the picture got noticeably more hopeful. A large-scale analysis by SchoolAI, examining more than 23,000 teacher-built AI activities from the 2024-25 school year, found that when teachers deliberately design and guide the interaction, AI is mostly being used to deepen critical thinking rather than replace it — 73% of these lessons required genuine conceptual understanding, 59% asked students to analyse information, and 58% asked them to evaluate ideas or form a judgement, not just retrieve a fact. The distinction the researchers kept coming back to wasn’t whether AI was present in a lesson. It was whether a teacher had actually shaped how students were allowed to use it.

For students who struggle with a particular subject, that shaped version of AI can genuinely function like a patient, always-available tutor — one that explains a concept a fourth different way without ever losing patience, and gives instant feedback a teacher juggling thirty students simply can’t provide in real time.

The Real Risk Isn’t the Tool — It’s the Shortcut
Strip away the disagreement, and both sides of this debate are actually worried about the same failure mode, just from different angles. If AI ends up doing the writing, the reasoning, and the analysis on a student’s behalf, the student loses the repetitions that would have built those muscles in the first place. Academic research on this is beginning to back up the intuition directly — a 2024 study using Bloom’s taxonomy on postgraduate business students found AI assistance genuinely helped with lower-order tasks like recall, but showed far less benefit, and real risk, for higher-order thinking — the analysis, synthesis and evaluation skills a degree is actually supposed to build. Offload enough of that over enough years, and a generation risks graduating with information at their fingertips but without much practised ability to use it unassisted.

Neither Banning Nor Ignoring It Is Working
What almost nobody studying this seriously recommends is prohibition. Banning AI outright hasn’t proven realistic anywhere it’s been tried, and treating it as a free-for-all clearly isn’t working either — the NEU’s finding that two-thirds of schools still have no student-specific AI policy, essentially unchanged from a year earlier, looks less like caution and more like institutional paralysis.

The more consistent recommendation, across researchers, teachers and policy groups alike, is somewhere in the middle: teach students to use AI deliberately and transparently, rather than either forbidding it or leaving it unsupervised. RAND’s researchers specifically point to models like the flipped classroom, where students engage with new material — AI-assisted or not — at home, and then do the actual thinking, discussion and practice during AI-free class time with a teacher present. Whatever the specific model, the researchers are unusually aligned on one point: schools need to tell students clearly and consistently when AI use is appropriate and when it isn’t, because a lot of the anxiety students report isn’t really about the technology — it’s about not knowing whether what they’re doing is even allowed.

AI, in other words, isn’t the enemy of learning, and it isn’t a replacement for it either. It’s a tool whose effect depends almost entirely on the hand guiding it — and right now, in most classrooms, that guidance still hasn’t caught up to how many students already have the tool open in another tab.

Share on Facebook

Post on X

Follow us

Save