Short excerpt: Artificial intelligence can become a powerful learning partner, but it should deepen a student’s thinking—not replace the difficult work through which understanding is built.
Every generation has worried that a new technology would make young people less capable. Calculators, the internet and smartphones all produced that anxiety. Artificial intelligence is different in one important way: it can now produce an answer, an argument and even a polished assignment in seconds.
That ability is impressive, but it creates a serious educational question. If a student receives a finished answer before learning how to frame the problem, check the evidence or explain the reasoning, has technology supported education—or quietly bypassed it?
What the latest PISA results tell us
The OECD published PISA 2025 Results: Future-Ready Students on 8 September 2026. More than 760,000 15-year-olds across 91 countries and economies took part. This is a broad international assessment, not a report about India, and its results should be read in that context.
The findings nevertheless deserve attention everywhere. Across OECD countries, average mathematics scores fell by 22 points between 2015 and 2025, while reading scores fell by 28 points. The report also found that 46% of students in OECD countries use AI chatbots at least weekly to help them learn.
Most importantly, the relationship between AI and performance was not simple. Students who regularly used AI for general purposes and had opportunities to develop AI literacy at school tended to perform slightly better in science. But students using AI for schoolwork tasks such as drafting, summarising or research recorded lower science scores than non-users in those comparisons. The OECD describes an association, not proof that AI itself caused the difference.
The danger is not the tool—it is intellectual dependence
I am not against students using AI. They will enter workplaces where these tools are normal, and refusing to teach them would be unrealistic. The real danger is allowing convenience to arrive before competence.
A young person must first learn to read patiently, write clearly, calculate, compare sources and defend a conclusion. These are not old-fashioned exercises. They are the foundations that allow someone to recognise when an AI answer is incomplete, biased or simply wrong.
In business, a confident presentation is never enough. We ask what assumptions support it, whether the numbers reconcile and what risks have been ignored. Students need the same habit. AI can make a weak idea look polished; education must teach them to look beneath the polish.
Use AI as a coach, not a substitute
The best use of AI is not “do this for me.” It is “help me understand this better.” A student can ask for another explanation of a difficult concept, generate practice questions, test an argument from the opposite side or receive feedback on a draft already written independently.
The sequence matters. Think first. Attempt the work. Use the tool to challenge or improve it. Then verify the output against reliable sources. When AI is used this way, it can encourage curiosity and give students more personalised support. When it supplies the first and final answer, it can remove the productive struggle through which learning happens.
Schools may therefore need to assess more than the submitted document. Oral explanations, supervised problem-solving, project work and questions about a student’s reasoning can show whether genuine understanding exists. The objective should not be to catch students using technology; it should be to make thinking visible.
Teachers and parents still matter most
UNESCO’s Guidance for Generative AI in Education and Research, originally published on 7 September 2023 and updated on 16 January 2026, calls for a human-centred, safe, equitable and meaningful approach. That is the right direction.
No technology can replace the teacher who notices that a child has lost confidence, or the parent who asks what the child actually understood. The OECD also reports that stronger relationships with teachers and families, supportive environments and a sense of belonging are connected with engagement and achievement.
This human layer becomes more important, not less, as tools become more capable. Students need adults who set boundaries, encourage effort and show that not knowing something is the beginning of learning—not an embarrassment to be hidden behind an instant answer.
Access must not create a new divide
AI can widen opportunity by giving more students access to explanations, language support and practice. But it can also widen inequality. The OECD notes that school opportunities to build AI literacy are more common among socio-economically advantaged students.
Giving everyone access to a chatbot is not the same as giving everyone the ability to use it wisely. Digital access must be accompanied by trained teachers, strong reading skills, source evaluation, privacy awareness and age-appropriate safeguards. Otherwise, the students who already have better support will extract more value, while others become more dependent on unverified answers.
My conclusion
AI should absolutely have a place in education, but the goal of education cannot be faster completion. It must remain deeper understanding, independent judgment and the confidence to solve unfamiliar problems.
My view is simple: teach students the fundamentals, then teach them how to use AI to extend those fundamentals. Let technology remove repetitive work, but never allow it to remove curiosity, effort or accountability.
The students who succeed in the AI era will not be those who can generate the quickest answer. They will be those who know which questions to ask, which answers to doubt and when to think for themselves.
Sources and context
Current figures and findings were checked against the OECD’s PISA 2025 Results, Volume I, published on 8 September 2026. Policy context was checked against UNESCO’s guidance on generative AI in education, last updated on 16 January 2026. Interpretations and recommendations in this article are my personal perspective.
— Tushar Kumar