Students already use AI for schoolwork. What PISA 2025 tells teachers
PISA 2025 is the first cycle to measure how students use AI chatbots. What the data says about AI use, performance and the risk of a new AI divide.
Key takeaways
- Students who do not use AI for specific schoolwork tasks, such as drafting writing assignments, outperform those who do.
- Weekly users and non-users of AI for general purposes perform similarly.
- Frequent users who are taught to assess AI-generated information tend to perform better than frequent users without that guidance.
- AI literacy opportunities are more common among advantaged students, which risks a growing AI divide.
- These are associations, not proof of cause, but they give schools a clear direction: purposeful use plus explicit AI literacy.
Students are not waiting for policy to catch up. In PISA 2025, students in most education systems reported using AI chatbots for schoolwork. In Denmark, for example, 48% of students used chatbots to help them learn at least weekly, close to the OECD average of 46%, and only 14% of students across the OECD reported never or almost never using chatbots for the schoolwork tasks PISA asked about.
What PISA found about AI and performance
The OECD calls the relationship between AI use and performance complex. Three findings matter most for teachers:
- Task matters. Students who do not use AI for specific schoolwork tasks outperform those who do. The OECD highlights drafting text for writing assignments: students who did not use AI for this outperformed those who did.
- General use is neutral. Weekly users and non-users of AI for general purposes perform similarly.
- Guidance matters. Among frequent users, students who used AI to help them learn and were taught how to assess the quality of AI-generated information tended to perform better than those who received no such guidance.
PISA’s dashboard now includes an indicator for “learning to assess AI-generated content in class”: the share of students who say they assess the quality of information produced by AI chatbots in their lessons at least sometimes.
“Technology and AI can strengthen learning … only when used purposefully and not as a substitute for attention, effort and understanding.”
Read the findings carefully
PISA is a survey, so these findings show associations rather than causes. Students who struggle with writing may be more likely to turn to AI for drafts, for example, rather than AI causing weaker writing. What the data does support is a practical distinction: AI that replaces thinking is a warning sign, while AI used alongside explicit instruction in evaluation is linked with better outcomes.
The emerging AI divide
The OECD found that opportunities to develop AI literacy are more common among socio-economically advantaged students. If some students learn to question, verify and improve AI output while others learn only to copy it, AI could widen existing gaps. Explicit AI literacy teaching in every classroom is an equity issue, not an optional extra.
Five classroom moves
- Think first, then AI. Ask students to attempt a plan, draft or solution before using AI, then use AI for feedback or comparison.
- Make evaluation routine. Build a short “check, compare, cite” step into any task where students use AI: check claims, compare with a trusted source and cite what they used.
- Separate learning tasks from product tasks. Where the goal is to build a skill, such as structuring an argument, keep AI out of the core step. Where the goal is exploration, AI can widen the range of ideas.
- Ask for process evidence. Short reflections, drafts and oral explanations show understanding in a way a finished product cannot.
- Teach it to everyone. Plan AI literacy lessons for the whole class rather than relying on students’ experiences at home. Our AI literacy activities are a starting point.
For a shared framework to organise this work, see the OECD–European Commission AI literacy framework explained.
Frequently asked questions
Does using AI lower students’ PISA scores?
PISA shows associations, not causes. Students who used AI for specific tasks such as drafting writing assignments scored lower than those who did not, while general weekly use showed no difference. Frequent users who were taught to evaluate AI output tended to perform better.
What is the AI divide in education?
The OECD found that opportunities to develop AI literacy are more common among advantaged students. That could widen achievement gaps if some students learn to use AI critically while others do not.
How should teachers respond to students using AI?
Teach AI literacy explicitly, ask students to think before using AI, build evaluation of AI output into tasks, and gather evidence of the learning process as well as the final product.
Sources
- OECD (2026), PISA 2025 Results (Volume I): Executive summary ↗
- OECD press release, 8 September 2026 ↗
- OECD (2026), PISA 2025 Denmark country note ↗
Figures are adapted from OECD (2026), PISA 2025 Results (Volume I): Future-Ready Students, OECD Publishing, Paris, https://doi.org/10.1787/73451bc5-en, published under CC BY 4.0. This is an adaptation of an original work by the OECD. The opinions expressed and arguments employed in this adaptation should not be reported as representing the official views of the OECD or of its Member countries. SparkSchool AI is not affiliated with or endorsed by the OECD.