Using Generative Artificial Intelligence for Reformatting References in Scholarly Writing

Marilyn H. Oermann, PhD, RN, ANEF, FAAN

Writer’s Camp Counselor

DeAnna Gapp, PhD, RN

Writer’s Camp Guest Counselor*

---

Our advice: Don’t try to use ChatGPT to format your reference list.

---

Generative artificial intelligence (genAI) such as ChatGPT and other AI tools are transforming nursing scholarship and scholarly writing. Authors are using genAI to brainstorm ideas for research and new practice and educational initiatives, analyze data, draft content for manuscripts, produce tables and figures, and complete other research tasks.^1-3 Issues with accuracy, relevance, bias, and plagiarism of the content generated by AI tools, however, have led to concerns about their reliability for scholarly writing.

Sources of genAI Information

GenAI can quickly create content for a manuscript. However, that content may not be accurate, relevant, or comprehensive. AI generates content using the information it was trained on, which changes daily. While this information includes peer reviewed articles that are open access, and were evaluated for accuracy and quality by the journal in which published, AI-created content also includes other information from the Internet: books, conference proceedings, and websites, which are generally not peer reviewed. The accuracy of that information is not known. GenAI also may miss key content from peer reviewed articles that are not open access and for which publishers have not authorized their use for AI training.⁴ Those articles, though, may be critical to understanding the topic and identifying gaps in the research. When AI tools are used to create content for manu­scripts, there also is the potential for bias (racial, gender, age, and others), for example, referring to nurses as females and physicians as males.

Accuracy of AI Generated Content

Many studies have documented errors in citations (in the text) and references (on the list at the end of the manuscript) including the digital object identifiers (DOIs) where accuracy is critical to locate a document. Mugaanyi et al.⁵ evaluated the accuracy of citations and references in an introduction to a manuscript created by ChatGPT (GPT-3.5). Half of the manuscripts were of topics in the sciences and the other half in the humanities. Only 73% of the citations in the sciences and 77% in the humanities were actual references. Many of the references created by genAI had errors in the DOIs. In another study by Kacena, Plotkin, and Fehrenbacher⁶ on whether ChatGPT could be used to assist in a research review and writing a credible review article, the content produced by ChatGPT had errors in references.

What does this mean for you as an author when considering the use of AI for your scholarship and scholarly writing? You are responsible for ensuring the content is accurate, relevant, comprehensive, and without bias. GenAI tools can assist in creating an initial draft, but we do not recommend this. Be aware that if AI is used for a draft or for revising the content, analyzing data, developing tables and figures, and completing other tasks that affect the content, you must disclose this in the manuscript and to the editor.^9-14 You are responsible for reporting how you used AI in your research and manuscript. Even with full disclosure, use of AI in this way might result in rejection by the editor if it is against journal policy.

Appropriate uses of genAI might include brainstorming ideas and potential research or clinical questions to guide a study, developing an outline for a manuscript, generating preliminary keywords for a search, translating text from another language into English, checking grammar and punctuation, and for general editing.⁸ Using tools to check grammar and spelling and to format references does not need to be disclosed.¹⁵

Formatting References

We wondered if GenAI would save us time in revising the format of citations and references to a different reference style. Many authors use reference management software (RMS) such as EndNote™, PaperPile, and Zotero to save citations during a literature search, retrieve and insert them into the text, and prepare a reference list. Another advantage of using RMS is that these programs can format the citations and references to multiple different reference styles, including APA¹⁶ and AMA,¹⁷ which are the most common styles used in nursing journals.¹⁸ Even though this seems like magic, you still need to check the formatting of both the citations and references—errors can occur due to how the data are entered in the RMS which affects the output. We decided to evaluate if genAI could be used to reformat the citations and references to a different reference style, a task often needed to submit a manuscript. If you are not using RMS, then you will need to do this manually. Could ChatGPT help?

We ran 6 tests using ChatGPT Version: GPT-4o (4 on 2025-06-23, conversion time between 16:00 EDT and 16:15 EDT, and 2 on 2025-06-25, conversion time 20:10 EDT). Table 1 provides the prompts and summaries of the results.

Test 1: In the first test, we started with a manuscript prepared using APA style. We initially confirmed there were no errors in the citations or references. We then asked ChatGPT to change the citations and references to AMA style.

Test 2: Because there were so many errors when changing from APA to AMA style in Test 1, we questioned if our prompt should have had American Medical Association instead of AMA. We modified the prompt to: Using the attached manuscript, change the citations and references to the American Medical Association reference style.

Test 3: In the third test, we started with a manuscript prepared using AMA style and first confirmed there were no errors in the citations and references. We asked ChatGPT to change the citations and references to APA style.

Test 4: We then modified the prompt to: Using the attached manuscript, change the citations and references to the American Psychological Association reference style.

Test 5: In our fifth test, we asked ChatGPT to change APA into Vancouver style, which is the same as the National Library of Medicine style. Citations are represented by numbers based on the order they are cited in the text, with the same number if used again.¹⁹ Citation numbers are in superscripts, parentheses, or brackets. References are numbered based on the order in the text (similar to AMA) and are not in alphabetical order.

Test 6: In the final test, we started with our manuscript in AMA style and asked ChatGPT to change the format to Vancouver style, which is similar.

Table 1. Converting Reference Formats

Test Number	Prompt	Summary of Results
1. Change APA to AMA	Using the attached manuscript, change the citations and references to AMA style.	With this prompt, ChatGPT left the format of the citations in APA style, added numbers to each paragraph and heading in the text, and continued with that numbering for the references. The references were still in APA style and in alphabetical order (instead of the order cited in the text as in AMA style). ChatGPT also deleted the DOIs.
2. Change APA to AMA	Using the attached manuscript, change the citations and references to American Medical Association reference style.	ChatGPT left citations that were integrated in a sentence in APA format. For example, both the original APA-formatted and AMA-converted manuscript had this same sentence: Ng et al. (2024) surveyed medical researchers…. When there were multiple citations in parentheses, however, ChatGPT replaced them with question marks. References were still in APA style and alphabetical order. ChatGPT deleted all of the DOIs except for one reference. Changing the prompt did not improve the accuracy of reformatting the citations and references.
3. Change AMA to APA	Using the attached manuscript, change the citations and references to APA style.	With this prompt, the text still included numbered citations (instead of author name and date); the references were numbered based on when they were cited in the text (instead of alphabetical order); ChatGPT added the year of publication in parentheses at the beginning of each reference; and the journal name was still abbreviated (AMA style).
4. Change AMA to APA	Using the attached manuscript, change the citations and references to the American Psychological Association reference style.	ChatGPT left the numbered citations in the text and on the reference list (original AMA style). The references were not changed to APA style. For most of them, ChatGPT added the year of publication in parentheses following the reference number and asterisks with the article title and journal name. Changing the prompt did not improve the accuracy of reformatting the citations and references.
5. APA to Vancouver style	Using the attached manuscript, change the citations and references to Vancouver style.	The citations were still in APA style instead of a numbered format. ChatGPT produced 2 separate reference lists for the manuscript. In the first list, ChatGPT added numbers in brackets to the references, still in alphabetical order, followed by a publication date in parentheses; removed author names from the references; kept the full name of the journal (APA style) but deleted the italics; and removed most of the DOIs. In the second reference list, ChatGPT added numbers to the alphabetized list of references instead of based on the order in which cited in the text, and the references were still in APA style. However, the journal names were no longer in italics, and ChatGPT removed most of the DOIs.
6. AMA to Vancouver style	Using the attached manuscript, change the citations and references to Vancouver style.	Citation numbers were changed from superscript to brackets, consistent with Vancouver style. However, ChatGPT added brackets to every number in the text including n’s, percents, and dates, for example, [27]%, n = [84], and in [2024]. Numbering of the references was not changed, but brackets were added to the reference numbers and to every number in the reference including the DOI, making the original reference impossible to locate.

ChatGPT was not useful in changing reference styles, and it introduced many errors. We might have improved accuracy if we modified the prompts further or tried a different AI tool. Our conclusion is this: it is easier to revise the reference style yourself if you need to change from one style to another and to do this, you should use RMS, such as Endnote, Paperpile, or Zotero. 

Conclusion

You can use AI tools to assist you in completing some of tasks associated with scholarly writing but not for changing the reference style. Regardless of the RMS you use, you need to review the citations and references and revise as needed to be consistent with the reference style of the journal. It is easier to make these changes in your reference library or manually rather than using a genAI tool. Our parting message is this: in writing a manuscript, use AI tools cautiously. You are the expert, and genAI cannot replace your expertise.

References

1. Van Noorden R, Perkel JM. AI and science: what 1,600 researchers think. Nature. 2023;621(7980):672-675. doi:10.1038/d41586-023-02980-0
2. Ng JY, Maduranayagam SG, Suthakar N, et al. Attitudes and perceptions of medical researchers towards the use of artificial intelligence chatbots in the scientific process: an international cross-sectional survey. Lancet Digit Health. 2025; 7(1):e94-e102. doi:10.1016/S2589-7500(24)00202-4
3. Khalifa M, Mona A. Using artificial intelligence in academic writing and research: An essential productivity tool. Computer Methods and Programs in Biomedicine Update. 2024:100145. doi:https://doi.org/10.1016/j.cmpbup.2024.100145
4. Schonfeld RC. Tracking the licensing of scholarly content to LLMs. The Scholarly Kitchen blog. October 15, 2024. https://scholarlykitchen.sspnet.org/2024/10/15/licensing-scholarly-content-llms/
5. Mugaanyi J, Cai L, Cheng S, Lu C, Huang J. Evaluation of Large Language Model performance and reliability for citations and references in scholarly writing: cross-disciplinary study. J Med Internet Res. 2024;26:e52935. doi:10.2196/52935
6. Kacena MA, Plotkin LI, Fehrenbacher JC. The use of artificial intelligence in writing scientific review articles. Curr Osteoporos Rep. 2024;22(1):115-121. doi:10.1007/s11914-023-00852-0
7. Ge L, Agrawal R, Singer M, et al. Leveraging artificial intelligence to enhance systematic reviews in health research: advanced tools and challenges. Syst Rev. 2024;13(1):269. doi:10.1186/s13643-024-02682-2
8. Oermann MH. You cannot search the literature using artificial intelligence, and this is why. Nurs Educ Perspect. 2024;45(6):337. doi:10.1097/01.Nep.0000000000001344
9. Zielinski C, Winker MA, Aggarwal R, et al. Chatbots, generative AI, and scholarly manuscripts: WAME recommendations on chatbots and generative artificial intelligence in relation to scholarly publications. Curr Med Res Opin. 2024;40(1):11-13. doi:10.1080/03007995.2023.2286102
10. International Committee of Medical Journal Editors. Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals. 2024. Accessed June 23, 2025. https://www.icmje.org/icmje-recommendations.pdf
11. Kaebnick GE, Magnus DC, Kao A, et al. Editors’ statement on the responsible use of generative AI technologies in scholarly journal publishing. Med Health Care Philos. 2023;26(4):499-503. doi:10.1007/s11019-023-10176-6
12. Hosseini M, Rasmussen LM, Resnik DB. Using AI to write scholarly publications. Account Res. 2023:1-9. doi:10.1080/08989621.2023.2168535
13. Oermann MH. Using AI to write scholarly articles in nursing. Nurse Educ. 2024;49(1):52-52. doi:10.1097/NNE.0000000000001577
14. Flanagin A, Pirracchio R, Khera R, Berkwits M, Hswen Y, Bibbins-Domingo K. Reporting use of AI in research and scholarly publication—JAMA Network guidance. JAMA. 2024;331(13):1096-1098. doi:10.1001/jama.2024.3471
15. Flanagin A, Kendall-Taylor J, Bibbins-Domingo K. Guidance for authors, peer reviewers, and editors on use of AI, language models, and chatbots. JAMA. 2023;330(8):702-703. doi:10.1001/jama.2023.12500
16. American Psychological Association. Publication manual of the American Psychological Association. 7th ed. 2020.
17. American Medical Association. AMA manual of style. 11 ed. Oxford University Press; 2020.
18. Oermann MH, Nicoll LH, Chinn PL, Conklin JL, McCarty M, Amarasekara S. Quality of Author Guidelines in Nursing Journals. J Nurs Scholarsh. 2018;297:1062. doi:10.1111/jnu.12383
19. Patrias K. Citing medicine: the NLM style guide for authors, editors, and publishers (2nd ed). 2015. Wendling DL, technical editor. National Library of Medicine (US); 2007 [updated 2015 Oct 2]. Available from: http://www.nlm.nih.gov/citingmedicine

Authors: Marilyn Oermann and DeAnna Gapp

Reviewed and Edited by Leslie H. Nicoll

Copyright © 2025 Writer’s Camp, Marilyn Oermann, and DeAnna Gapp. CC-BY-ND 4.0

Citation: Oermann M, Gapp DA. Using Generative Artificial Intelligence for Reformatting References in Scholarly Writing. The Writer’s Camp Journal, 2025; 1(2):12. doi:10.5281/zenodo.16813876

*DeAnna Gapp, PhD, RN teaches nursing at Washtenaw Community College in Ann Arbor, Michigan. With experience teaching at several universities, she finds particular fulfillment in working with community college students, whose dedication and resilience continually inspire her. As a first generation college student, her journey in nursing began with a degree from Monroe County Community College, finished by a Ph.D. in Educational Studies with a focus on Nursing Education from Eastern Michigan University. DeAnna frequently publishes and presents on topics in nursing education, including artificial intelligence. ORCiD: https://orcid.org/0000-0002-8093-2451