Study Finds AI Use Raised Homework Grades But Lowered Exam Scores

A new study found that students using AI scored better on homework but worse on exams, suggesting reduced learning retention.

Students who used AI assistance on homework saw their assignment scores rise — but when it came time for closed-book exams, their performance dropped, according to a study circulating via The Economist and posted to SSRN, a preprint research repository. The findings point to a pattern where AI may help complete tasks without building the underlying knowledge those tasks are meant to develop. The research appears to quantify a concern educators have raised since AI tools became widely accessible: that offloading cognitive work to AI produces short-term grade gains while undermining long-term learning. The study has been linked to SSRN, meaning it is likely a working paper and may not yet have undergone full peer review.

Why it matters

If AI assistance inflates homework grades while depressing exam performance, it raises serious questions about how schools measure learning and whether AI access is widening gaps between assessed and actual student knowledge.

What's next

The paper's peer review status and methodology details are not yet clear from available sources; further scrutiny of the study's design and sample size will determine how broadly its conclusions apply.

Key facts

• Students using AI on homework saw homework scores increase
• The same students showed lower exam scores compared to a baseline
• The study was posted to SSRN, a preprint academic repository
• The Economist covered the study, lending it mainstream visibility
• The findings suggest AI assistance may substitute for learning rather than supplement it

Bias & framing notes

Only one source (The Economist) contained any substantive reporting, and the body text was not directly available — only archive and SSRN links were provided. The second source (canews24.online) had no body text at all. Without access to the study's methodology, sample size, or peer review status, core claims cannot be independently verified. The score reflects the credibility of The Economist as an outlet and the existence of a linked primary paper, offset by the lack of accessible detail.

NewsClear — neutral news & congressional tracking · Bill of the Week