Journal retracts study claiming ChatGPT improves student learning

In short: Springer Nature has retracted a study that claimed ChatGPT improves student learning, after finding problems in how the results were analyzed.

What happened

The journal Humanities & Social Sciences Communications, published by Springer Nature, has retracted a paper that reported strong benefits from using OpenAI’s ChatGPT in education. The retraction notice was posted on April 22, 2026, almost a year after the study first appeared on May 6, 2025.

The study was a “meta-analysis,” which means it tried to combine results from many earlier studies (like averaging many report cards to get one overall grade). It reviewed 51 prior research papers and compared groups of students who used ChatGPT with groups who did not. The authors concluded that ChatGPT had a “large positive impact” on learning performance, plus moderate benefits for students’ perception of learning and for “higher-order thinking” (more complex thinking, like evaluating and reasoning).

Springer Nature said it found “discrepancies” in the meta-analysis and that these issues weakened confidence in the paper’s conclusions. The publisher also said the authors did not respond to messages about the retraction.

Critics, including University of Edinburgh lecturer Ben Williamson, said the paper appeared to mix studies that were not truly comparable. He also questioned whether enough high-quality classroom research on ChatGPT could have been completed so soon after ChatGPT’s release in November 2022.

Why it matters

The paper had already spread widely before it was retracted. It received hundreds of citations, including 262 citations in Springer Nature journals and 504 citations overall, and it attracted major attention online. Retractions often do not travel as far as the original headlines, so the claim that “ChatGPT helps learning” may continue to be repeated even after the study was pulled.

Source: Arstechnica