AI didn’t kill academic publishing. It exposed the body

When the bibliography of an academic article contains fake authors and non-existing papers, something is clearly wrong. Citations are a form of academic credit for labor and intellectual contribution, and they play an important role in how recognition and legitimacy is distributed in academia. 

In a recent piece for the Dutch weekly De Groene Amsterdammer, Utrecht University PhD candidate Joris Veerbeek looked at 136,000 academic publications in the Netherlands to identify papers with AI-hallucinated references. The findings paint a rather grim picture about the future (and past) of the academic paper.

Not just an educational problem
Since the launch of ChatGPT, every university teacher has noted a dramatic change in student submissions and participation. Writing essays, completing assignments, and even reading papers are now often executed with the help of AI chatbots. Universities were quick to lament the decline of academic education and responded with inconsistent policies that simultaneously restrict and encourage the use of AI tools. Much of this response has centered on students: how they use GenAI, when its use should be allowed, what transparency requirements are needed, and how assignments can be made “AI-proof.” 

GenAI has exposed how educational systems have come to value product over process. Yet the implications of GenAI for academic research remain largely overlooked. It is as if issues such as plagiarism and academic fraud are seen as the exclusive domain of students rather than academics themselves. 

In the meanwhile, submissions to journals have become inflationary, with an increase of up to 50% for some disciplines. The avalanche of AI-supported papers makes it hard for editors to manage the influx of papers and for reviewers to evaluate them. Yet for unclear reasons, the academic paper and its associated H-index remain the dominant measures of academic achievement and continue to shape career paths. 

A much deeper crisis
The findings of Veerbeek and colleagues reveal symptoms of a much deeper and structural problem. Indeed, long before the advent of GenAI, academic publishing was already marked by serious problems. Poorly written papers and sloppy research could pass through peer-review; minimally altered versions of the same article could appear in multiple outlets; and full professors could add their names to papers they barely glimpsed at. 

The billion-dollar industry of so-called paper mills and predatory journals further exposes the scale of the problem. Citation cartels compounded these issues, as researchers inflated citations to themselves and their networks, while some reviewers pressured authors to cite their work before recommending publication. An unbearable workload combined with persistent budget cuts did the rest. 

The ease and quality of academic publication has not only been affected, we also witness an epistemological crisis unfolding. Citations no longer reliably serve as foundations for accounts of knowledge production. As more papers containing hallucinated references find their way into repositories and discourse, trust in academic work will crumble further. This crisis is not entirely new. Online discussions about hallucinated citations reveal that the fast pace of academic work has long encouraged researchers to cite articles they have only skimmed, often relying on little more than the abstract. This pressure has contributed to calls for Slow Academia, a movement against the neo-liberal, metric-driven university.

Unsustainable
The academic publishing system was already cracking. Algorithmic writing machines are merely the straw that breaks its back. Academic institutions now largely ignore the problem, deflect it by focusing on student misconduct, or pursue futile efforts to identify AI-generated slop automatically (spoiler alert: this is not effective). They do so rather than recognizing the need for more tangible formats of knowledge transfer and better indicators for academic labor, impact and excellence. So where do we go from here? University administrations, research associations, and full professors must reckon with the unsustainable paper production and the failures of its quality-assurance systems. They need to embrace alternative forms of knowledge transfer and broader criteria for assessing academic relevance and excellence. 

We see an opportunity to strengthen academic inquiry beyond the traditional paper publication and the confines of the university. We can engage with novel but effective forms of knowledge transfer such as action research, collaborative and transdisciplinary problem-solving, and novel educational formats.

Mirko Tobias Schäfer, Sciences Lead Data School
Karin van Es, Humanities Lead Data School