Large university study finds ChatGPT had no detectable effect on grades after COVID disruption was modeled

emollick · x · 2026-07-25

A large university study found that once COVID-era disruption was modeled separately, introducing ChatGPT had no detectable effect on grades. Course evaluations for subject understanding, interest, and relative workload also showed no meaningful change.

The study tested the substitution hypothesis versus the idea that GenAI helps learning. Using administrative data from one of the largest U.S. universities, the authors compared susceptible assessments such as take-home essays with less susceptible ones across the pre- and post-ChatGPT periods.

Original post →

More from Research

Research channel →