AI Grades Student Essays More Generously Than Professors, Study Finds

Universities around the world are facing mounting pressure on their teaching staff, and artificial intelligence has frequently been proposed as a solution to ease the burden of grading written assignments. A recent study, however, suggests this solution may have serious limitations.

According to Inside Higher Ed, researchers found that large language models (LLMs) tend to award higher grades to student essays compared to human professors. The gap is not minor but systematic, raising questions about the accuracy of these tools when used for formal academic assessments.

Why AI's leniency matters

Using AI to grade essays seems appealing at first glance: professors would save time, and students would receive feedback faster. But if AI-generated grades are consistently higher than those given by humans, they no longer accurately reflect students' actual level of performance.

Researchers cited by the outlet stress that AI models "cannot be considered a reliable indicator" of genuine academic performance — a crucial point for institutions considering partial or full automation of the grading process.

What this means for universities

The study, authored by Juliette Rowsell for Times Higher Education, arrives at a time when many higher education institutions are actively exploring AI solutions to reduce faculty workload amid growing class sizes and limited resources.

The research findings suggest that while AI can be useful as a supporting tool — for instance, providing quick initial feedback or flagging obvious errors — it should not fully replace human judgment in final evaluations, especially where grades have direct consequences for students' academic paths.

As more universities experiment with such technologies, the study brings back into focus a central dilemma of digital education: how to harness AI's efficiency without compromising the fairness and rigor of academic assessment.