66% of students now revise with an AI chatbot — two in three of them. Not as a future possibility, but as a normal, everyday study habit, right now.
That is the headline finding of Digital Learning: Insights from the Classroom, a new report from Oxford University Press. Oxford surveyed 600 secondary school students and 400 teachers across four countries — South Africa, Brazil, the United Kingdom and Hong Kong — with focus groups in the UK on top of that. And it is very clear about the direction of travel.
What the report actually found
66% of learners said they used AI chatbots to help them prepare for exams. Students described AI as an interactive study aid — something they can talk to, ask follow-up questions of, and use as a stand-in guide when a teacher is not around to answer. Free tools make that easy: ChatGPT and Google Gemini in an app or browser, and Meta's chatbots inside WhatsApp and Facebook, which use less mobile data.
And teachers? Teachers use chatbots even more than their students do:
- 82% of teachers use them for research
- 74% to make learning materials
- 66% for admin work
- 56% for marking
- Only 3% said they never use AI chatbots at all
Read those two sets of numbers together and the old framing falls apart. This is not "students sneaking a new tool past teachers who don't use it." Both sides of the classroom are already using the same tools, heavily, for ordinary work.
The part that gives me hope
The finding I keep coming back to is not a percentage. It is the line students drew themselves.
Most learners in the survey said using chatbots to revise, plan and spark ideas was fair — but using them to do the work for them defeated the point of school. One learner put it better than most policy documents do: "People aren't retaining the information if they're just copying and pasting — it kind of defeats the point of learning."
That matters, because it means we are not starting from zero. Students are not, as a group, demanding the right to outsource their learning. Many of them already sense the difference between help me understand this and do this instead of me. Our job is to make that instinct explicit, practised and consistent — before the shortcut version becomes the habit.
The two honest complications
Access is not equal. In South Africa, 18% of students surveyed reported having no access to a digital device in at least one setting — compared with 13% in the UK, 4% in Brazil, and none in Hong Kong — even though 91% of South African teachers surveyed had laptops. When revision help lives inside a chatbot, a device gap quietly becomes a learning gap. Any school AI plan that ignores access is only a plan for the students who are already equipped.
Distraction is real. 52% of teachers and 47% of students identified distraction as the biggest challenge of digital learning. The same device that quizzes a student at 9 p.m. also carries every other app they own. Oxford's own conclusion is worth copying here: digital integration is not a binary choice between digital and analogue — teachers and students are combining both, intentionally, based on what works. Pedagogy first, format second.
What this changes in my classroom
I teach ICT and Robotics in Bangkok, and I already teach AI use openly rather than banning it. This report reinforces three moves I think every teacher can make, whatever their subject:
1. Teach the line, don't just police it. Name the difference out loud, the way the students in this survey did: AI for revising, questioning, planning and checking is learning support; AI producing the work you submit is substitution. My classroom test stays the same — could you explain and defend this work if I asked you about it right now?
2. Design revision tasks that use the chatbot well. If two in three students revise with AI anyway, I would rather set the good version as homework: "Ask the chatbot to quiz you on today's topic, bring me the two questions you got wrong, and show me how you checked its answers." That turns a private habit into a taught skill — including the verifying, because chatbots sound confident even when they are wrong.
3. Declare the help. In my classes, if AI helped with any part of a task, students write one line saying what they asked and what they changed. Declaring help is never punished; hiding it is the problem. The Oxford numbers tell me this needs to be normal school culture now, not one teacher's experiment — teachers are using these tools for research, materials, admin and marking, so modelling honest use is something we owe students, not just demand from them.
The honest caveat
This is a survey, not a census: 600 students and 400 teachers in four countries, describing their own habits. Oxford itself is careful to call it a snapshot. So treat the exact percentages as a strong signal about direction rather than universal facts — the numbers in your school will differ. The pattern, though, is hard to argue with: on both sides of the classroom, these tools are already ordinary.
The real takeaway
Banning AI was always going to be a losing battle, but this report shows something more useful than that: the battle framing itself is out of date. Students are using chatbots to study, teachers are using them to teach, and both groups — in their own words — mostly want the use to stay honest and purposeful.
So the real job now is not keeping AI out of revision. It is teaching students to use it to think harder: to revise, to question, and to check. That is a teaching job, and it is ours.
How do you draw the line in your classroom between "AI helped me learn" and "AI did it for me"?
Sources: Oxford University Press, Digital Learning: Insights from the Classroom — report overview; reporting via Tzaneen Voice / MyBroadband
Title image: original clay artwork created for this blog.
