WORLD - As millions of students return to classrooms for a new academic year, artificial intelligence is becoming an increasingly familiar part of school life. A question that once seemed hypothetical is now confronting teachers, parents and students: if an AI chatbot can complete an assignment in seconds, what is the purpose of doing the homework at all?
The debate is no longer simply about whether students should be allowed to use AI. It is about whether schools need to rethink what they consider learning, and how they measure it.
AI can make education more accessible. A student struggling with a mathematical concept can ask for another explanation. Someone learning a new language can receive instant feedback. A difficult text can be broken down into simpler terms.
But the same technology can also remove the struggle that learning often requires.
Better Homework, Worse Exam Results?
That distinction is becoming clearer as researchers examine what happens when students use generative AI for schoolwork.
A large study involving around 27,000 students aged 12 to 18 in China offers one of the clearest warnings so far. Researchers from Stockholm University and the University of Hong Kong followed students for 30 months, comparing their homework performance with their results in exams where AI was not available.
The contrast was striking.
After students began using generative AI, their homework scores increased by 18 percent. At the same time, they spent roughly 30 percent less time completing each assignment, with the average falling from 64 minutes to 45 minutes.
On the surface, this looks like an educational success story: students were completing better work in less time.
The problem appeared when the technology disappeared.
Within six months, students' scores on monthly closed-book exams had fallen by 20 percent. Their performance in high-stakes entrance examinations also declined, although the full effect took longer to emerge.
The findings point to a gap between producing an answer and acquiring knowledge.
The Difference Between Assistance and Replacement
Homework has traditionally been more than a way for teachers to check whether students can produce the correct response. It allows students to retrieve information, practise skills, make mistakes and work through problems independently.
When AI performs those steps instead, students may still submit better-looking work without developing the abilities that the work was supposed to build.
The Organisation for Economic Co-operation and Development (OECD )has highlighted a similar concern in its Digital Education Outlook 2026. Its review of emerging research found that when generative AI is simply used to outsource tasks, students can perform better without necessarily experiencing meaningful learning gains.
This is where researchers have raised concerns about what has been described as “metacognitive laziness”: a tendency to rely on technology instead of engaging in the mental effort required to understand, evaluate, and remember information.
Yet the problem may not be AI itself. How students use it matters.
Not All AI Use Is the Same
A survey of 7,000 students aged 12 to 17 across seven European countries found that 56 percent use AI to find information and 45 percent use it to explain terms or concepts. These uses can potentially support learning by helping students understand material they find difficult.
But 31 percent said they use AI to provide complete solutions to tasks.
That difference is crucial.
Asking an AI system to explain why an answer is correct requires a student to engage with the material. Asking it to produce the answer removes much of that process.
The challenge for schools, therefore, may not be to keep AI outside the classroom, but to teach students how to use it without allowing it to replace their own thinking.
Schools Are Already Adapting
Some education systems are already moving in that direction. China has introduced mandatory AI instruction, while countries in the Gulf, including the United Arab Emirates, Saudi Arabia and Qatar, have begun incorporating AI into education.
The question is whether these efforts will focus simply on teaching students how to use AI tools, or also on teaching them when not to use them.
That distinction could become increasingly important as AI becomes more capable and more deeply integrated into everyday education.
Beyond Homework
The transformation is also extending beyond assignments and exams.
A 2026 survey of more than 1,000 university students in the United Kingdom found that 49 percent believed AI had improved their student experience. Students cited saving time, improving their understanding and receiving immediate, personalised support.
AI is also beginning to occupy a more personal space. When asked about loneliness, 59 percent said AI had made no difference, while 21 percent felt less lonely and 20 percent felt more lonely.
That suggests that the role of AI in education could eventually extend far beyond homework. It may become a tutor, study partner, organiser and, for some students, even a source of conversation.
Keeping the Thinking Human
The central challenge is therefore not simply preventing students from using AI. It is preserving the parts of education that cannot be outsourced.
Schools may need to place greater emphasis on closed-book assessments, classroom discussion, oral presentations, problem-solving and projects that require students to demonstrate how they reached an answer — not simply whether they have one.
AI can make learning faster. But education has never been only about speed.
If students become better at completing schoolwork while becoming worse at thinking independently, then higher homework scores may conceal a deeper educational problem.
The real test of AI in education may ultimately be simple: when the chatbot is switched off, what can the student still do?
