New Delhi: The use of artificial intelligence (AI) in submissions of scientific papers to a major journal has jumped up 42 per cent since November 2022, when ChatGPT was first released to the public, while quality in writing has declined, according to a study.
The study looked at nearly 7,000 abstracts submitted by 11,887 authors to the academic journal Organization Science -- considered a premiere publication in management and organisation theory -- between January 2021 and February 2026 that were reviewed about 10,400 times by more than 2,500 reviewers. Over 8,000 peer reviews were also analysed for the use of AI.
The journal is the first to report early impacts of AI usage in the review process of scientific publishing, researchers, including those from The Wharton School at the University of Pennsylvania and Duke University, said.
AI-based writing is flooding the journal with a high volume of low-quality papers, and the surge has strained the volunteer labour that evaluates research, which might partly explain the high percentage of reviews with AI-generated signatures we observe, the team said.
"AI is placing the peer-review system under stress that shows no signs of decreasing," they said.
They added that AI is not levelling the playing field and that all author teams from non-native english institutions and new field entrants who expect AI to strengthen writing receive little benefit in terms of an increased chance of publication.
Further, AI reviews are not equivalent to or better than human reviews -- they are less well-written and narrower in systematic ways.
"As the AI Task Force for Organization Science, we provide an early account of artificial intelligence's (AI) impact on both submissions and reviews at a major academic journal," they wrote, and added that what they observed is not limited to the journal or social sciences.
"Submission volume has risen 42 per cent since the late 2022 release of ChatGPT, while writing quality has declined. The rise in AI-generated writing accounts for nearly all of these trends," they said.
The researchers said that while they cannot assess appropriate or ideal levels of AI usage, they can conclude that "the current state of AI tools, amplified by existing publish-or-perish incentives, appears to be pushing the system toward an equilibrium of more rather than better research."
Each text was analysed by the tool 'Pangram', a deep learning model trained to classify a text as human or AI-generated. The tool is said to report a 99.84 per cent accuracy on human-written text.
By February 2026, majority of submissions to the Organization Science journal used AI in writing to some degree, with "the most striking" trend seen in the rise of reports where 70 per cent or more of the text was generated by AI, the researchers found.
The team said that a heavy reliance on AI for writing manuscripts is not solely due to the introduction of large language models themselves but instead, appears to reflect authors' "rational response" to their institution's incentives, particularly at schools which emphasise journal lists and publication counts.
This report was published from a syndicated wire feed. Apart from the headline, the EdexLive Desk has not edited the copy.
