Commentary from the Camp Director: Reference Managers Redux

Reference Managers, Artificial Intelligence, and the Integrity of Scholarly Writing

Leslie H. Nicoll, PhD, MBA, RN, FAAN

Writer’s Camp Director

---

Now more than ever, a reference manager must be part of your writing armamentarium.

---

Seeing nurse-authored work published in a major medical journal such as The Lancet carries a particular kind of significance. It feels a little like breaking the white-coat ceiling: nursing scholarship crossing into spaces historically dominated by medicine.

That was my reaction recently when I read correspondence by Max Topaz and colleagues addressing fabricated references in the biomedical literature.¹ Their analysis, based on an audit of 2.5 million biomedical papers, describes a growing problem driven in part by generative artificial intelligence (AI) writing tools that generate plausible-sounding but nonexistent citations. Some fabricated references contain real author names, convincing article titles, credible publication years, and believable journal information—yet the cited articles do not actually exist.

The correspondence appropriately emphasizes publisher workflows, automated verification systems, integrity metadata, and post-publication correction mechanisms. These approaches will likely become increasingly important as submission volumes grow and generative AI-assisted writing tools become more common. Yet as I read the article, I found myself thinking about an important omission: the role of the author as the first integrity checkpoint in scholarly communication. Topaz et al. neglect the author in both identifying the problem and proposing possible solutions.¹ A subsequent editorial in The Lancet likewise emphasized that authors bear responsibility for the entire contents of a manuscript, including references, and noted that fabricated citations raise broader concerns about research integrity and accountability.²

The way I see it, the deeper issue is not simply fabricated references. It is the gradual weakening of scholarly contact with sources.

For more than 30 years, I have encouraged authors to use reference managers such as Pro-Cite, EndNote, and more recently, Paperpile.^3–8 In the 1990s and early 2000s, my rationale was straightforward: bibliographic management software reduced typographical errors, simplified formatting, saved time during manuscript revision, and helped writers organize growing collections of scholarly literature. At the time, reference managers were often viewed primarily as productivity tools.

Today, however, the stakes are much higher. Reference managers are not simply formatting tools. They are part of the integrity infrastructure of scholarly publishing.

Reference Errors Are Not New

Concerns about reference accuracy long predate generative AI. In 1987, Foreman and Kirchhoff documented inaccuracies in references published in nursing journals, highlighting a problem that editors and authors have wrestled with for decades.⁹ Subsequent studies, including my own work examining citation accuracy in nursing journal articles, demonstrate that errors in author names, titles, page numbers, and publication details remain common.^10,11

Even fabricated references are not entirely new. Cases of research misconduct and retractions have occasionally involved invented or unverifiable citations long before generative AI became widely available.^12–14 What is different now is the speed, scale, plausibility, and ease with which entirely fabricated references can be generated by generative AI systems.

Traditionally, citation inaccuracies occurred within references to real publications. The current concern about hallucinated citations represents something fundamentally different because the underlying source may not exist at all.

Scholarship Requires Contact with Sources

At its core, scholarship depends on engagement with sources. A citation is not simply a string of formatted text added to satisfy a journal requirement. It represents an encounter between the author and the scholarly literature: the process of locating a source, reading it, evaluating its relevance, interpreting its findings or arguments, and deciding how it contributes to the manuscript being written.

For generations, this workflow created a built-in integrity checkpoint. Before citing a paper, an author generally had to obtain the article itself. In the print era, this might involve a trip to the library stacks, an interlibrary loan request, or a mailed reprint card. Later, electronic databases and PDFs accelerated access dramatically, but the underlying expectation remained the same: scholars cited works they had actually consulted.

Artificial intelligence changes this workflow in important ways. Large language models can generate references that look entirely plausible, even when no underlying article exists. More importantly, fabricated citations can enter a manuscript without the author ever encountering a source at all. The traditional scholarly sequence—find, read, evaluate, cite—can be bypassed entirely.

Frankly, I struggle to understand how a hallucinated reference makes its way into a manuscript in the first place. As the first author on anything I publish, I have read—or at least carefully skimmed—every source included in the reference list. That expectation may sound old-fashioned, but it reflects what I believe remains a core principle of scholarship: authors should maintain direct contact with the literature they cite.

This is one reason fabricated references feel so unsettling to many editors and scholars. The problem is not simply that a citation is incorrect. A fabricated citation represents something different: a reference untethered from any actual scholarly object and disconnected from direct author engagement with the literature.

What Reference Managers Actually Do

Most authors think of bibliographic software primarily as a formatting tool: something that inserts citations into a manuscript and generates a properly styled reference list. Although these functions are useful, they are only part of the story.

Reference managers are also metadata managers. When an author imports a citation directly from MEDLINE/PubMed, a journal website, Ovid, or another indexed source, the software captures structured metadata: author names, article title, journal title, DOI, publication year, volume, issue, page numbers, and often abstracts or PDF links. When used correctly, bibliographic management software preserves information imported from authoritative databases and connects the citation to an identifiable scholarly object.

In other words, the citation enters the manuscript through a documented pathway connected to a real publication. Its provenance is preserved.

This is why I have spent decades encouraging students, authors, faculty members, and editors to use reference managers. My reasons were never limited to saving time or simplifying citation styling. Bibliographic management systems improve accuracy, preserve source information, and encourage disciplined scholarly habits. The process is not complicated: find the source, read it, add it to your reference manager, and verify the details (remember the old computer adage: GIGO, garbage in, garbage out). Browser add-ons for popular programs such as Paperpile or Endnote make this as easy as a few mouse clicks. Scholars really have no reason not to use one of these systems. The common excuses of “too hard to learn,” or “too time consuming,” no longer pass muster. The research by Topaz et al. confirms that.¹

Bypassing this workflow creates problems. Manually typing references into a database, copying citations from old manuscripts, or pasting references into a bibliography manager or manuscript breaks the connection between the citation and the underlying source. A fabricated or incorrect citation entering the database effectively disappears: the software will format it perfectly and may never signal that anything is wrong. A reference manager cannot fix input errors or verify a source that never existed in the first place.

Fabricated versus Predatory References 

Fabricated references are not the only threat to reference integrity. Nursing editors have also spent years sounding the alarm about citations to predatory or deceptive journals.^15–17 Unlike hallucinated references, predatory articles do exist. Authors can retrieve them, read them, and cite them—sometimes knowingly and sometimes without recognizing the nature of the publication.

This distinction matters because the two problems require different responses.

Hallucinated references are, at least theoretically, detectable through technical verification systems. If an article does not exist in any indexed database, DOI registry, publisher archive, or bibliographic record, software may eventually be able to identify the discrepancy automatically.

Predatory references are harder. The article may exist, possess a DOI, appear in Google Scholar or PubMed Central, and resemble a legitimate publication. Determining whether the source is credible, appropriate, or trustworthy requires scholarly judgment, disciplinary literacy, and author knowledge.

Future reference-checking systems will likely become increasingly sophisticated. I can imagine automated tools that verify DOI metadata, identify retracted articles, flag suspicious journal records, or detect nonexistent citations before submission. These developments will be important and necessary. But no automated system can fully replace scholarly understanding of the literature itself.

Responsibility Starts with Authors

The correspondence by Topaz and colleagues¹ focuses primarily on publisher systems, screening tools, integrity metadata, and post-publication correction mechanisms. These are important interventions. However, their discussion pays comparatively little attention to the author as both the origin point of the problem and the most scalable point of prevention.

Editors, peer reviewers, publishers, and indexing services all play important roles in maintaining the integrity of the scholarly record. However, when it comes to references, the first and most important checkpoint remains the author. This is partly an issue of responsibility, but it is also an issue of scale.

An author preparing a manuscript may need to verify 25, 50, or perhaps 100 references. Although this requires time and attention, it is a finite and manageable task. The author has direct access to the original sources, understands the context in which the references are being used, and is in the best position to confirm that each citation is accurate and appropriate.

Editors operate in an entirely different environment. A journal editor may oversee hundreds or thousands of manuscript submissions each year, many containing dozens of references. Even a modest-sized journal can process tens of thousands of citations annually. Under these conditions, comprehensive manual reference verification becomes extraordinarily difficult.

Peer reviewers face similar constraints. Reviewers are asked to evaluate scientific merit, methodology, interpretation, originality, organization, and relevance, usually under significant time pressure and without compensation. Although reviewers may notice suspicious citations, they cannot realistically authenticate every reference in every manuscript they review.

This is why integrity at the point of authorship matters so much. When authors use reference managers correctly, they reduce errors before the manuscript ever reaches the editor or peer review process. Prevention during manuscript preparation scales much more effectively than retrospective editorial checking later in the publication process. It is far easier for one author to verify 50 references in one manuscript than for one editor to verify 50,000 references across an entire journal system.

Conclusion

Reference managers have long been presented as productivity tools—useful for organizing articles, formatting citations, and saving time during manuscript preparation. Those functions still matter. But in the age of AI-generated text and fabricated references, bibliographic management software must also be understood as part of the integrity infrastructure of scholarly publishing.

A properly maintained reference library does more than produce correctly formatted citations. It preserves metadata, maintains connections to real scholarly objects, and supports accurate attribution and verification.

Authors should know where their references came from, what those sources say, and why they are being cited. A reference list is not decorative formatting added at the end of a manuscript; it is part of the evidentiary structure of scholarly writing.

For more than 30 years, I have encouraged writers to use reference managers because they improve efficiency, organization, and accuracy. Today, I am making the argument even more strongly. In an era when fabricated references can be generated instantly and made to look convincing, disciplined bibliographic workflow is no longer simply a matter of convenience. It is part of responsible scholarly practice.

Use a reference manager. Build your library carefully. Import citations from authoritative sources. Preserve the metadata.

Be an accountable author.

References

1. Topaz M, Roguin N, Gupta P, Zhang Z, Peltonen LM. Fabricated citations: an audit across 2·5 million biomedical papers. Lancet. 2026;407(10541):1779-1781. doi:10.1016/s0140-6736(26)00603-3
2. Bauchner H, Rivara FP. Fabricated references: a new threat to editorial integrity. Lancet. 2026;407(10541):1765-1766. doi:10.1016/s0140-6736(26)00798-1
3. Nicoll LH. Finding and documenting sources. Saver C, Ed. Anatomy of Writing for Publication for Nurses. Sigma Theta Tau Publishing; 2011; 2014; 2017; 2021 (4 editions)
4. Nicoll LH. EndNote version 4 (software review). Computers in Nursing. 2001;19(1):6-10.
5. Nicoll LH, Ouellette TH, Bird DC, Harper J, Kelley J. Bibliography database managers. A comparative review. Computers in Nursing. 1996;14(1):45-56. Accessed May 8, 2026. https://www.ncbi.nlm.nih.gov/pubmed/8605660
6. Nicoll LH. Managing bibliographic information. The Journal of Nursing Administration. 1993;23(9):13-14. doi:10.1097/00005110-199309000-00005
7. Nicoll LH. A practical way to create a library in a bibliography database manager: using electronic sources to make it easy. CIN: Computers, Informatics, Nursing. 2003;21(1):48-54. doi:10.1097/00024665-200301000-00014
8. Nicoll LH. Camp Director Commentary: Artificial intelligence and me. The Writer’s Camp Journal, 2025; 1(3):9. doi:10.5281/zenodo.17450174
9. Foreman MD, Kirchhoff KT. Accuracy of references in nursing journals. Research in Nursing & Health. 1987;10(3):177-183. doi:10.1002/nur.4770100310
10. Nicoll LH, Oermann MH, Carter-Templeton H, Wrigley J, Owens JK. Exploring the accuracy of cited references in a selected data set of nursing journal articles. ANS Advances in Nursing Science. 2022;45(3):209-217. doi:10.1097/ANS.0000000000000408
11. Nicoll LH, Oermann MH, Chinn PL, Conklin JL, Amarasekara S, McCarty M. Guidance provided to authors on citing and formatting references in nursing journals. Journal for Nurses in Professional Development. 2018;34(2):54-59. doi:10.1097/NND.0000000000000430
12. Retraction. “Results of psychometric testing of the RADS-2 with school-based adolescents seeking assistance for sexual orientation and gender identity concerns. Part 2: Research brief”: Retraction. Journal of Child and Adolescent Psychiatric Nursing: Official Publication of the Association of Child and Adolescent Psychiatric Nurses, Inc. 2011;24(3):196. doi:10.1111/j.1744-6171.2011.00284.x
13. Retraction. “Depressive illness in teens and preteens and effectiveness of the RADS-2 as a first-stage assessment. Part 1: Descriptive paper”: Retraction. Journal of Child and Adolescent Psychiatric Nursing: Official Publication of the Association of Child and Adolescent Psychiatric Nurses, Inc. 2011;24(3):196. doi:10.1111/j.1744-6171.2011.00283.x
14. Retraction. “Treatment of trauma- and abuse-related dissociative symptom disorders in children and adolescents”: Retraction. Journal of Child and Adolescent Psychiatric Nursing: Official Publication of the Association of Child and Adolescent Psychiatric Nurses, Inc. 2011;24(3):197. doi:10.1111/j.1744-6171.2011.00285.x
15. INANE “Predatory Publishing Practices Collaborative”. Predatory publishing: What editors need to know. Nurse Author & Editor. 2014;24(3):1-5. doi:10.1111/j.1750-4910.2014.tb00183.x
16. Oermann MH, Conklin JL, Nicoll LH, et al. Study of predatory open access nursing journals. Journal of Nursing Scholarship. 2016;48(6):624-632. doi:10.1111/jnu.12248
17. Oermann MH, Nicoll LH, Ashton KS, et al. Analysis of citation patterns and impact of predatory sources in the nursing literature. Journal of Nursing Scholarship. 2020;52(3):311-319. doi:10.1111/jnu.12557

Author: Leslie H. Nicoll

Reviewed and Edited by: Marilyn Oermann

Copyright © 2026 Writer’s Camp and Leslie H. Nicoll CC-BY-ND 4.0

Citation: Nicoll LH. Reference managers, artificial intelligence, and the integrity of scholarly writing. The Writer’s Camp Journal, 2026; 2(2):8. doi:10.5281/zenodo.20101699