Opening Pandora’s box: paper mills in conference proceedings

Keywords: paper mills, conference proceedings, collaboration anomalies, co-authorship

• Paper mills are a threat to research integrity.

• The study investigates the penetration of paper mills into conference proceedings, focusing on IEEE, which is one of the top publishers in terms of retractions worldwide.

• By analysing more than 4,000 offers for sale available on different social media platforms, we identified 1,720 papers associated with paper mills and co-authored by more than 6,500 authors from 55 countries.

• Conference proceedings should have more rigorous peer review and quality control.

Introduction

Paper mills, commercial entities that sell co-authorship slots and violate the peer-review process, are an increasing threat to research integrity (Richardson et al., 2025). Although the phenomenon was documented more than a decade ago (Hvistendahl, 2013), our knowledge about paper mills is still fragmented by discipline or geography (Abalkina et al., 2025). The majority of studies look at the impact of paper mills on journals (Van Noorden, 2023); there is still a gap in the literature regarding the penetration of paper mills in conference proceedings. In some fields, particularly in computer science, conference proceedings are the primary venue for original research results. The goal of this study is to identify paper mill papers in the conference proceedings of IEEE, which is one of the biggest organisers of conferences and the top publisher in terms of retractions worldwide (Lendvai & Sasvári, 2025), and analyse scientometric and content-based patterns of such paper-mill papers.

Data and methods

One of the methods used by paper mills is offering titles for sale on social media or online websites. We have identified over 200 social media channels on Telegram, Instagram, Facebook, etc., which (re-)post offers for sale. From July 2025 to January 2026, we searched for titles of conference papers sold on various social media channels and published by 31 December 2025. The identified offers were posted between February 2021 and December 2025. We collected over 4,000 unique paper titles.

An offer for a conference paper typically includes the conference publisher, the title of the paper, the offer’s publication date, and the available slots for purchase. We searched for published papers matching these offers both manually, using the search functions in Google Scholar, Scopus, and IEEE Xplore, and semi-automatically, using Crossref data. The latter was done by comparing the offered titles to the titles of all recently published IEEE conference papers, using either the Levenshtein distance (ignoring capitalization, punctuation, and spaces) or cosine similarity between extracted Sentence-BERT embeddings (Reimers & Gurevych, 2019), and then manually examining the most similar matches. Each paper has been verified by two authors who classified papers as “definitely”, “probably”, and “no match”. Papers receiving “no match” were excluded from the final sample, while those receiving “probably” were discussed further.

Findings

We have identified 1,752 unique IEEE conference papers that initially matched the collected titles. The final sample comprised 1720 papers because 32 papers were excluded due to ‘no match’ or as a result of our discussion. These papers are co-authored by more than 6,500 researchers affiliated with over 3,500 institutions in 55 countries. 93% of authors represent India, 17% - USA, 11% - Iraq.

The identified papers have been published in 286 conferences, accounting for up to 23,51% in an individual conference.

The identified papers show several anomalous collaboration patterns. First, 82.8% of papers have exactly six co-authors, compared with only 15.6% of all conference papers published by the IEEE in 2024. Since 2020, the number of IEEE-published conferences with mostly six-author papers has increased in contrast to conferences with other authorship sizes, suggesting that paper mills have systematically infiltrated many such venues. Second, the identified papers have a high diversity of affiliations per paper (5.1 on average), a characteristic commonly associated with paper-mill publications (Abalkina, 2023). Third, 43% of the papers include authors from at least two countries, indicating that paper-mill operations are international in scope. Fourth, although the primary clients of paper mills are authors from low- and lower-middle-income countries, there is evidence that authors from high-income countries also participate in dishonest collaboration with paper mills.

We also identified 139 papers linked to authorship offers in Springer Nature, Taylor & Francis, AIP Publishing, Elsevier, EDP Sciences, and three other conferences.

Conclusions

Our study demonstrated the infiltration of paper mill papers into conference proceedings, affecting more than 280 conferences published by IEEE and other conference providers. Conferences can be particularly vulnerable to paper mill penetration due to less rigorous and time-limited peer review, as well as light-touch control over the organizers.

This study has limitations. The majority of offers have been published in English, which biases the results toward certain countries. The overall number of paper mill papers is much higher because offers are often deleted, titles can be changed before publication, not all offers are public, and we probably didn’t identify all social media channels.

References

Abalkina, A. (2023). Publication and collaboration anomalies in academic papers originating from a paper mill: Evidence from a Russia‐based paper mill. Learned Publishing, 36(4), 689-702.

Abalkina, A., Aquarius, R., Bik, E., Bimler, D., Bishop, D., Byrne, J., ... & Wise, N. (2025). ‘Stamp out paper mills’—science sleuths on how to fight fake research. Nature, 637(8048), 1047-1050.

Hvistendahl, M. (2013). China`s publication bazaar. Science.

Lendvai, G. F., & Sasvári, P. ‘Wasted’ research and lost citations: A scientometric assessment of retracted documents in Scopus between 2001 and 2024. Journal of Information Science.

Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Sentence embeddings using Siamese BERT-networks. In K. Inui, J. Jiang, V. Ng, & X. Wan (Eds.), Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) (pp. 3982–3992). Association for Computational Linguistics. https://doi.org/10.18653/v1/D19-1410

Richardson, R. A., Hong, S. S., Byrne, J. A., Stoeger, T., & Amaral, L. A. N. (2025). The entities enabling scientific fraud at scale are large, resilient, and growing rapidly. Proceedings of the National Academy of Sciences, 122(32), e2420092122

Van Noorden, R. (2023). How big is science’s fake-paper problem. Nature, 623(7987), 466-467.