Students who use AI generally score worse at school
Source Entity
Dominic Preston
A recent OECD PISA report indicates that students who frequently use AI tools generally perform worse academically than their non-AI-using peers. However, the data suggests that students trained to critically evaluate AI outputs may see performance benefits, highlighting the need for better digital literacy in education.
The Impact of AI on Global Academic Performance
A recent report from the OECD’s Programme for International Student Assessment (PISA) has ignited a critical conversation regarding the integration of artificial intelligence in educational environments. Based on data from 2025—the first major assessment cycle since AI tools achieved mainstream adoption—the study evaluated over 760,000 students across 91 countries. The findings reveal a troubling trend: students who rely on AI to assist with their studies generally perform worse in core subjects like science, mathematics, and reading compared to those who do not use these tools.
Nuance in the Data: The Critical Thinking Variable
While the headline data suggests a negative correlation between AI usage and academic outcomes, the reality is significantly more nuanced. The report highlights that the detrimental effects are not universal. When the data is adjusted for socioeconomic factors, the performance gap persists, yet a distinct subgroup of learners shows different results. Students who are specifically taught to critically assess the reliability and accuracy of AI outputs demonstrate a slight performance boost. This suggests that the issue is not the technology itself, but the methodology of its application.
The Shift in Educational Paradigms
This PISA report serves as a landmark moment in pedagogical research, marking the first time global educational standards have had to account for the pervasive presence of generative AI. Historically, educational systems have struggled to adapt to technological shifts, such as the rise of the internet or personal calculators. However, AI presents a unique challenge because it can simulate human-level reasoning and synthesis, potentially bypassing the cognitive "struggle" that is essential for deep learning and knowledge retention.
Implications for Global Literacy Standards
The findings underscore a widening divide between passive AI consumption and active, critical engagement. If students use AI merely as a shortcut for homework completion, they forfeit the opportunity to build foundational problem-solving skills. Conversely, those who utilize AI as a collaborative tutor—by verifying its claims and questioning its logic—can potentially leverage the technology to enhance their understanding. This transition requires a fundamental shift in classroom instruction, moving away from rote memorization toward a focus on digital information literacy.
Future Trends and Policy Considerations
Looking ahead, educational systems worldwide will likely need to formalize AI literacy as a core competency. The PISA data suggests that without structured guidance, students are prone to cognitive atrophy when using AI. Policymakers and educators must now decide whether to integrate AI tools into formal curricula with strict oversight or implement restrictive policies. The evidence suggests that a total ban is unlikely to be effective; rather, the focus must shift toward training students to treat AI as a fallible assistant rather than an authoritative source of truth.
Conclusion
The 2025 PISA report provides a sobering baseline for the digital age. While the immediate correlation between AI and lower test scores is a cause for concern, the potential for positive outcomes through critical assessment offers a path forward. The challenge for the next decade will be to bridge the gap between AI accessibility and academic integrity, ensuring that technology serves as a bridge to knowledge rather than a barrier to cognitive development.