KQ1: What are the features of paper mill products?

Knowledge of the features of paper mill products enables their detection and deterrence. The largely incidental discoveries of publications from paper mills, combined with limited coordinated research across stakeholder groups, have likely produced incomplete and fragmented knowledge [9]. We need more complete and contemporary knowledge of both discipline-specific and -agnostic features of paper mill products to improve detection and deterrence.

KQ2. What is the scale of the paper mill problem?

Once features of paper mill support are known, these features must be leveraged to inform understanding of the scale of paper mill-supported publications, across fields, topics, methodologies, and over time. Scaled temporal literature analyses can allow the discovery of new features, further informing KQ1, and answer questions such as when paper mills likely commenced using generative AI, and how their product features are likely to change [9].

KQ3. How do paper mills operate and evade detection?

While some features of paper mill operations could be predicted by models of operational rationality [12], we currently know very little about paper mill operations. Large, well-resourced paper mills could conceivably provide more sophisticated manuscripts and services [1,5] and could represent arms of other unethical businesses [1,14]. As some information on paper mill operations is now dated [1,3,10], we need research to inform current paper mill operations in different countries and settings, and the demographics of paper mill employees and clients. We also need current information about how paper mills collude with researchers, institutions, journals, and/or publishers to scale operations and evade detection, including the possible use of generative AI and commercially available screening tools. Understanding how paper mills produce manuscripts could also valuably inform product features (KQ1) and problem scale (KQ2).

KQ4. To what extent are researchers and scholars aware of paper mills?

While publishers are increasingly aware of paper mills [2,11], little is known about levels of awareness within other stakeholder groups. As awareness can help researchers to recognize and avoid paper mill products [7], research is required to inform the levels of awareness within different research and scholarly communities. The resulting information must then inform the design and implementation of awareness-raising campaigns and education programs, informed by evidence from KQs 1–3, and KQ5 below, to support all researchers, including those serving as journal editors and peer reviewers [2].

KQ5. How are paper mills affecting science and scholarship?

Just as there is currently limited information about awareness, little is known about how paper mill products are impacting research and scholarship. For example, researchers could unknowingly cite paper mill products, potentially slowing research translation [9] or impacting patient care [4]. We therefore need to understand how paper mill products are cited in different fields and across different publication types and how these citations influence author, publication, and journal metrics. With improved awareness of paper mills (KQ4), researchers could also change their practices, problem choices or research directions in response to publications from paper mills, where individual responses might vary according to research field and/or career stage. At present, we have only scattered anecdotal descriptions (for example, from social media) of how some researchers have been affected or are choosing to respond.

In summary, paper mills represent a threat to genuine research and scholarship that requires expertise, resources, and time [9]. In any call for research on paper mills, it is important to recognize ongoing efforts to address this problem. For example, journals and publishers are leveraging known paper mill features to detect submissions and retract publications, where cross-publisher collaborations such as the STM Integrity Hub provide education, shared tools, and workflows [2,9,11]. At the same time, paper mill product features and operations could rapidly evolve in response to improved detection and new capabilities [2,9,11]. Publishers and researchers alike will therefore benefit from timely research on paper mill products and operations to improve both awareness and responses (Table 1).

We recognize that paper mills represent a challenging topic for researchers and funders alike. Nonetheless, research funders and institutions must now take courageous decisions to provide the necessary resources to transform our understanding of paper mills. Dedicated funding of paper mill research will also signal that paper mills and research fraud represent legitimate, important topics. Research support must enable rapid, ambitious research at scale, matched with systems that fast-track research translation to the literature and its many user communities (Table 1). For example, translation of paper mill research requires faster processes for achieving post-publication corrections at scale [9]. Knowing that papers can be quickly flagged where there is strong suspicion of mill involvement will also be a powerful motivator for researchers and funding agencies, who may likewise see little point in identifying problematic papers if these papers simply remain uncorrected.

Research in all fields relies upon the integrity of the literature. Paper mills and research misinformation undermine trust in research and should therefore be recognized as major global challenges, of similar importance to emerging pandemics or climate change. We must therefore waste no time in discovering the full extent of the paper mill problem and in taking all necessary steps to protect genuine science and scholarship.

Acknowledgments

The authors thank Sabina Alam (Taylor and Francis Group), Deborah Kahn (independent consultant), Hylke Koers (STM Solutions), Elizabeth Moylan (Wiley), Joris van Rossum (STM Solutions), and Alison Avenell (University of Aberdeen) for critical reading and discussions, and gratefully acknowledge support from United2Act, a project jointly funded by COPE and STM. We apologize to authors whose publications were not cited due to reference limits.