The Classroom Without a Textbook
What Saudi Arabia’s AI Bet Reveals About the Future of Learning
eyesonsaudi-arabia
In September 2026, a slightly different kind of school year began across Saudi Arabia. The International Center for AI Research and Ethics in Riyadh, working with Google, opened free year-long access to Google’s Gemini AI Plus plan to university students across the Kingdom — the latest expansion of an offer first piloted for Saudi and Egyptian students in late 2025, now scaled into a nationally coordinated enrollment drive rather than a self-serve sign-up page. It sits alongside something even more sweeping: since the 2025–2026 academic year, more than six million Saudi schoolchildren, from elementary grades through secondary school, have been studying an entirely new national subject — artificial intelligence itself, taught as a core part of the curriculum rather than an elective bolt-on. The UAE has moved on a nearly identical timeline. Between the two of them, the Gulf’s two largest economies have made AI literacy a baseline expectation for an entire generation within the space of about eighteen months.
It’s a useful anchor for a bigger question that schools everywhere are now wrestling with, whether or not their governments have announced a national AI curriculum: does learning with AI actually make students sharper — or does it just make them faster at looking like they’ve learned something?
The Case That AI Is Making Students Better Learners
The research base, still young but growing fast, is more favorable to AI than the popular anxiety around it suggests. A 2026 meta-analysis pooling 35 experimental studies on ChatGPT’s effect on student learning found a moderately positive overall effect on both cognitive and non-cognitive outcomes, with the strongest gains showing up where AI was woven into structured, blended instruction rather than used as a free-for-all. Other systematic reviews report similar patterns: students working with AI chatbots often show improved engagement, stronger self-directed learning habits, and — when the interaction is built around guided questioning rather than direct answers — measurable gains in critical thinking. One study of sixth-graders found significant improvements in both grades and information retention among students who used a chatbot compared with a traditional constructivist curriculum.
The mechanism researchers keep pointing to is not the AI’s intelligence but its patience: unlimited, judgment-free repetition, instant feedback, and content that adjusts to a student’s specific gaps rather than the pace of the whole classroom. For students who would otherwise sit quietly confused at the back of a lecture, that alone can be transformative.
The Case That Something Is Being Lost
Set against that is a 2025 MIT Media Lab study that has become the most-cited cautionary tale in this debate. Researchers split 54 participants into three groups — one writing essays entirely unaided, one using a search engine, one using ChatGPT — and tracked their brain activity with EEG throughout. The pattern was stark: brains of participants using ChatGPT showed significantly less activity and engagement than those working entirely unaided. By the third essay, many ChatGPT users were pasting the model’s output with only minor edits, and when the tool was later taken away, they struggled to accurately recall or quote from essays they had supposedly just written. The researchers called the pattern “cognitive debt” — the idea that today’s shortcut quietly withdraws from tomorrow’s capacity to think, recall, and create unassisted.
Other studies land somewhere in the messy middle. Some report no significant difference at all in learning outcomes between AI-assisted and traditionally taught groups, concluding that the technology itself matters less than how deliberately it’s integrated into teaching. A German study tracking university students’ performance on critical-reasoning tasks found that heavy chatbot use didn’t reliably predict better or worse results — it mostly reflected how the tool was used, not whether it was used.
So — Are AI Students “Smarter” Than Book Readers?
The honest answer is that the question is probably the wrong shape. The emerging evidence doesn’t point to AI users versus book readers as two competing tribes with one objectively ahead. It points to a difference between two modes of using any tool, AI included: outsourcing versus scaffolding.
A student who asks an AI model to explain a concept three different ways, then closes the laptop and tries to reconstruct the explanation from memory, is doing something closer to what a great tutor or a demanding book does — forcing the brain to do the retrieving, which is where real learning happens. A student who asks the same model to simply produce the essay is doing the cognitive equivalent of photocopying someone else’s homework, dressed up in a conversational interface. Traditional reading has never been immune to its own version of this problem — skimming a summary instead of the book is its own kind of shortcut — but the friction involved in reading a dense text has historically forced at least some deeper processing that a frictionless chat window doesn’t require unless a student deliberately builds it back in.
What the research does suggest fairly consistently is that AI’s biggest wins show up in motivation, engagement, and access — getting more students to actually start the work, ask questions they’d have been too embarrassed to raise in class, and stick with subjects they’d otherwise have given up on. Its biggest risks show up in exactly the skills that used to be built by the unglamorous grind of reading, rereading, and writing without a shortcut on hand: sustained attention, memory consolidation, and the sense of authorship over your own ideas.
What Saudi Arabia’s Bet Is Really Testing
That’s what makes the Gulf’s experiment worth watching closely over the next few years. Teaching AI as a subject — how it works, where it fails, what it’s for — is a genuinely different intervention from simply handing students unlimited access to a chatbot and hoping for the best. The explicit ambition behind Saudi Arabia’s curriculum is to produce a generation that treats AI as a tool to be understood and directed, not a shortcut to be leaned on unconsciously. Whether six million students actually end up on the “scaffolding” side of that line or the “outsourcing” side won’t be settled by the policy announcement — it will be settled classroom by classroom, teacher by teacher, in the small daily choice between letting a chatbot answer a question and asking it to help a student work the answer out themselves.
The printed book never had to worry about that distinction; it was incapable of doing your thinking for you even if you wanted it to. AI’s real novelty in education isn’t that it’s smarter than a textbook. It’s that, for the first time, the technology in a student’s hands is capable of thinking so that the student doesn’t have to — which makes the choice of how to use it matter more than it has for any learning tool before it.
Research findings cited above come from peer-reviewed studies and systematic reviews published between 2025 and 2026; this is an active and fast-moving area of research, and findings should be read as an emerging picture rather than a settled consensus.
