Why AI Might Be Making Students Score Lower

Since the explosive arrival of generative AI, a quiet debate has been raging in households and classrooms around the world. Is artificial intelligence the ultimate personalized tutor, capable of explaining complex physics concepts at 2 AM? Or is it a sophisticated crutch that encourages intellectual laziness?

We finally have some concrete data to help answer that question, and the initial results are sobering. According to the latest Programme for International Student Assessment (PISA) report by the OECD, students who rely on AI to help them study generally perform worse academically than those who avoid the technology altogether.

The PISA study, which evaluated 15-year-old students globally in science, math, and reading based on data collected in 2025, marks the first comprehensive look at global education since AI tools became a mainstream staple on students' smartphones and laptops. The headline finding seems to validate the worst fears of skeptical educators: integrating AI into a study routine correlates with lower test scores.

However, digging deeper into the data reveals a much more nuanced reality. The problem doesn't appear to be the technology itself, but rather the way young minds are interacting with it.

Learning is inherently a process of productive struggle. When a student wrestles with a difficult algebra equation or tries to parse the meaning of a dense historical text, new neural pathways are formed. If a student simply feeds a prompt into a chatbot and copies the beautifully synthesized answer, the assignment gets completed, but the learning process is entirely bypassed. The AI does the heavy lifting, leaving the student's cognitive muscles unexercised.

Yet, the OECD report highlights a fascinating exception to this downward trend. There is a specific subset of students who actually experience a slight academic boost from using AI. Who are they? They are the learners who have been explicitly taught to critically assess how well the AI tools perform.

When students treat AI not as an infallible oracle, but as a flawed assistant whose work must be double-checked, edited, and questioned, the dynamic changes completely. This active engagement—evaluating the machine's logic, spotting hallucinations, and refining the output—requires a high level of critical thinking. In this scenario, the AI becomes a sparring partner rather than a shortcut.

This global data provides a crucial roadmap for the future of education. Banning AI in schools is likely a futile endeavor; the technology is too accessible and too deeply integrated into modern life. Instead, the focus must shift toward teaching digital skepticism. The most valuable skill in the AI era won't be knowing how to write the perfect prompt, but knowing how to interrogate the answer.

Key Points

• The 2025 OECD PISA report found that 15-year-olds using AI to study generally score worse in math, science, and reading.
• Passive reliance on AI bypasses the 'productive struggle' necessary for genuine learning and cognitive development.
• A notable exception exists: students explicitly taught to critically evaluate AI outputs actually see a slight academic boost.
• The findings suggest schools should focus on teaching digital skepticism rather than attempting to ban AI tools.

Why It Matters

As AI becomes an inescapable part of the educational landscape, this global data proves that the mere presence of technology doesn't guarantee better learning—success depends entirely on active, critical human engagement.

Sources:

• Students who use AI generally score worse at school — The Verge - AI