Since the public launch of ChatGPT in November 2022, academic journals have experienced a massive increase in article submissions. But behind this apparent boom in scholarly productivity lies a sobering reality. New research from Lundquist College of Business Associate Professor of Management Alex Murray and his coauthors has found that the spike is driven almost entirely by the use of AI, and it's pushing the scientific community toward a crisis of quantity over quality.
In their groundbreaking paper published in April 2026, "More Versus Better: Artificial Intelligence, Incentives, and the Emerging Crisis in Peer Review," the research team—Murray, Claudine Gartenberg at the University of Pennsylvania, Sharique Hasan at Duke University, and Lamar Pierce at Washington University of St. Louis—secured approval to access and analyze the full corpus of manuscript submissions and reviews at Organization Science, regarded as one of the top journals for management, business, and related fields. (Murray and Gartenberg are senior editors, Hasan is a deputy editor, and Pierce is editor-in-chief of Organization Science, and they all serve on the journal's AI Task Force).
What they specifically uncovered is that the ease of generating text with large language models (LLMs) has fueled a staggering 42 percent increase in manuscript submissions to the journal. Yet, this increased volume is not translating into higher quality. Instead, the authors found that submissions that rely heavily on AI to generate text suffer from a marked decline in readability and are burdened with denser, more complex jargon.
This foundational shift in how academic research is being disseminated and evaluated has sparked feature coverage in Nature and Forbes, as well as interest from OpenAI, where Gartenberg presented the paper's findings.
During his presentation of the paper at the University of Oregon in May, Murray underscored that the project is not a blanket indictment of AI or LLMs.
"In this analysis, we do not take any normative stance on the inappropriate usage of AI," Murray explained. "This is about the system as a whole, the papers that are coming in, what's getting published, and where might we be going."
Murray recalled that the idea for the paper came during a meeting of journal editors in Hong Kong where they struggled to describe a new wave of "weightless, hard-to-parse manuscripts." He noted that scholar Joel Baum from the University of Toronto has colorfully described them as "research-shaped objects."
As Murray put it, "The artifact is a paper, kind of looks like a paper, but when you dive into it, you're like, what is this? What am I reading? It's dense, hard to parse through, you read it two times, you're like, wait, they're actually not saying anything at all."
Because of this drop in quality, these AI-generated manuscripts are facing a harsh reality at the editorial desk. Submissions with high levels of AI usage face desk rejection rates approaching 70 percent, rarely making it past the initial screening.
Furthermore, the paper notes that "the authors who are most likely to use AI in writing—research teams from non-native English-speaking institutions and new entrants to the field—do not benefit in the review process and, in fact, may be hurt when using AI for writing."
In addition, the flood of AI-generated submissions is placing a massive burden on the volunteer labor of journal editors and peer reviewers. Overwhelmed by the sheer volume of dense papers hitting their desks, exhausted reviewers are beginning to turn to AI to manage their own workloads. The study revealed that more than 30 percent of review reports submitted to the journal now show signs of AI usage as well.
Just like the manuscripts they evaluate, these AI-generated reviews tend to be harder to read and narrower in scope. Rather than rigorously scrutinizing data and empirical methods, AI reviews tend to focus heavily on theoretical framing and front-end presentation, Murray noted.
The result is an emerging ecosystem crisis where automated papers risk being evaluated by automated reviews.
As the paper warns, "If AI only amplifies volume—and lowers quality in the process—we may end up with more research but less collective knowledge."
And although AI is the tool accelerating this crisis, the coauthors emphasize that the technology isn't creating these flaws out of thin air.
"AI isn't introducing problems, it's exposing problems that are already deeply embedded and ingrained in the systems of knowledge production," Murray said. "If you're incentivized to publish more, why would you write one journal article that has five studies in it? Why not parcel them—see if you can get two or three papers out there. We're seeing now you have a tool that can make that more efficient."
As the study goes on to highlight, the surge in AI-generated manuscripts is directly correlated with universities that place a high premium on publication counts in elite journals, such as those in the UT-Dallas list. In some regions, the authors note, institutions offer staggering direct financial bonuses ranging from $20,000 to $60,000 per top-tier publication, heavily incentivizing researchers to prioritize volume and speed over rigorous, groundbreaking work.
Murray emphasized that the current trajectory of AI submissions is unsustainable and outlined a potential dystopian scenario in which submission volumes climb, elite editors withdraw due to burnout, and AI review agents contend with an ever-growing number of AI-generated submissions.
Avoiding this path will require deliberate, structural choices that extend far beyond journal editorial offices. The research team argues that university leadership must do the hard work of shifting rewards away from simple publication counts and back toward quality.
"If we can do this," Murray noted, "academia can reach a state of equilibrium where instead of mass-producing papers, AI is used to identify meaningful empirical puzzles, build richer, more robust theories, and conduct large-scale analyses that weren't previously possible."
"Reaching an equilibrium in which AI serves as a critical engine of innovation will require that our institutions and the incentive structures they create adapt," the paper concludes.
Until then, academic publishing risks idling for a long time in a phase where editors are constantly swatting away low-quality submissions at the expense of deeper scientific discovery and true knowledge advancement.
—Jim Engelhardt, Lundquist College Communications