The DOI matches its Crossref record; 29 of the 29 references it cites were found in a registry or an index; none of the 1 sources its claims cite is checked for support yet.
Hallucinated citations produced by generative artificial intelligence may constitute research misconduct when citations function as data in scholarly papers
Open accessAbstract
In this article, we discuss the growing problem of hallucinated citations produced by Generative Artificial Intelligence (GenAI) in scholarly research and writing. We argue that GenAI hallucinated citations might qualify as a provable instance of research misconduct under the U.S. federal regulations when a) the researcher uses a GenAI tool to produce hallucinated (i.e., nonexistent) citations for a research document; b) the citations function as data because they directly support research findings, as in, for example, review articles or bibliometric studies; and c) the researcher demonstrates indifference to the risk of fabrication of the data (i.e. citations) because they did not check the GenAI's output for veracity and accuracy. Other types of problematic citations such as bibliometrically incorrect citations, or contextually inaccurate citations, are indicative of poor scholarship and irresponsible behavior, but do not qualify as research misconduct. Recognizing that GenAI hallucinated citations could be regarded as research misconduct in certain cases will hopefully encourage researchers to take this problem more seriously than they do now. In partnership with scientific institutions, funders and professional societies, the scholarly community should work on establishing, promoting, and enforcing standards for responsible use of AI in research, including standards pertaining to citation practices.
Overview
CoChat Summary
TL;DR
This conceptual/legal analysis argues that GenAI-hallucinated citations may constitute provable research misconduct under U.S. federal regulations when citations function as data and the researcher recklessly fails to verify them. It distinguishes such cases from other problematic citations, which it treats as poor scholarship rather than misconduct.
Key findings
- The authors argue hallucinated citations may qualify as research misconduct when: (a) a researcher uses GenAI to produce nonexistent citations; (b) the citations function as data because they directly support research findings, as in review articles or bibliometric studies; and (c) the researcher shows indifference to the risk of fabrication by not checking GenAI output for veracity and accuracy.
- A hallucinated citation does not qualify as plagiarism, and the authors say it would not qualify as falsification because GenAI makes up a citation rather than changing an existing one; however, it could qualify as data fabrication when the citation is necessary to validate findings.
- Other problematic citations—bibliometrically incorrect citations or contextually inaccurate citations—are indicative of poor scholarship and irresponsible behavior but do not qualify as research misconduct.
- Reported examples include: 19 of 29 citations hallucinated in an *Academic Ethics* paper; 18 of 76 cited citations hallucinated in a retracted *PLOS ONE* study; 12 of 14 cited references hallucinated in a *Digestive Diseases and Sciences* perspective; and in a mental health study of GPT4o-generated citations (n = 176), 56% contained errors, with one in five hallucinated, though rates varied by disorder from 6% to 46%. A cited meta-analysis found almost 17% of quotations in medical publications have inaccuracies, with roughly half involving major errors.
Methods
- Design: normative/legal and conceptual analysis, not an empirical study.
- Data/sample: no original dataset; the authors analyze U.S. federal research misconduct regulations, including the OSTP 2000 definition of fabrication, falsification, and plagiarism and the 2024 Public Health Service definition of recklessness, alongside definitions of research data from OMB 1999 and EU 2019.
- Analysis: applies FFP and recklessness criteria to GenAI hallucinated citations, using hypothetical examples such as a systematic review of adverse drug reactions and a bibliometric study of authorship trends.
Limitations
- The authors state that treating hallucinated citations as data fabrication is a “novel (and perhaps even radical) idea” that may not square with preconceived ideas about fabricating data, and they acknowledge potential enforcement concerns for academic institutions.
- The authors state they are not proposing retrospective audits of the scientific literature and are not claiming every hallucinated citation qualifies as research misconduct; only cases where citations serve as data to support findings might qualify.
- Evident from the text: the argument relies on hypothetical examples and selected reported cases rather than systematic empirical evidence, and the list of problematic citation types is described as non-exhaustive.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 0 Not valid · 1 Not checked · 0 Own results and interpretations
- It may state two separate claims as one. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Published guidance warns against hallucinated references, privacy leaks, overreliance, and funder-policy violations [8,19].”
From A Workflow for Grant Discovery and Proposal Development Using Large Language Models: Development and Formative Evaluation (2026) · cites this paper as [19]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“A workflow without issue-based state tracking, deduplication memory, or citation audit checkpoints risks producing output that lacks a recoverable decision history, cannot be verified against eligibility criteria, and may include fabricated references that would only be detected by the funder [19].”
From A Workflow for Grant Discovery and Proposal Development Using Large Language Models: Development and Formative Evaluation (2026) · cites this paper as [19]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Finally, the large language model citation audit checkpoint applies to research outputs produced by the system, but grant proposals submitted to external funders require additional manual verification for every factual claim, since hallucinated citations in a grant application are a credibility risk that automated checks may not fully prevent [19].”
From A Workflow for Grant Discovery and Proposal Development Using Large Language Models: Development and Formative Evaluation (2026) · cites this paper as [19]
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Self-check results appear here once an independent check has confirmed them.
Cited works that were retracted
Works this paper cites that were later retracted, from the retraction notices publishers deposit. Major when the notice came out before this paper was published.
A note, not an issue: for example, a cited work that was retracted only after this paper was published, which its authors could not have known.
In the paper
“10.1371/journal.pone.0302484”
Cited work 10.1371/journal.pone.0302484 · retraction date not known · this paper published 2026-03-15
Citations
Cites
1 retracted
Editors, T. P. O. (2024) · PLoS ONE · doi:10.1371/journal.pone.0302484 · Cited by 4
RetractedHow to conduct a bibliometric analysis: An overview and guidelines
Donthu, N., Kumar, S., Mukherjee, D., Pandey, N., & Lim, W. M. (2021) · Journal of Business Research · doi:10.1016/j.jbusres.2021.04.070 · Cited by 14,329 · cited 1×
“Bibliometrics research involves identifying, classifying, mapping, and analyzing various references (Donthu et al. 2021).”
Resnik, D. B., & Hosseini, M. (2025) · AI and Ethics · doi:10.1007/s43681-024-00493-8 · Cited by 228
Self-citationResponsible Conduct of Research
Shamoo, A. E., & Resnik, D. B. (2022) · doi:10.1093/oso/9780197547090.001.0001 · Cited by 224
Self-citationAn International Study of Research Misconduct Policies
Resnik, D. B., Rasmussen, L. M., & Kissling, G. E. (2015) · Accountability in Research · doi:10.1080/08989621.2014.958218 · Cited by 185 · cited 2×
“Most of the misconduct definitions used in other countries are similar to the U.S. definition (Resnik, Rasmussen, and Kissling 2015).”
Self-citationResearch Misconduct Definitions Adopted by U.S. Research Institutions
Resnik, D. B., Neal, T., Raymond, A., & Kissling, G. E. (2014) · Accountability in Research · doi:10.1080/08989621.2014.891943 · Cited by 124 · cited 1×
“…definitions of misconduct used by U.S. universities or other organizations include misbehaviors other than FFP (e.g., serious violations of human or animal research regulations), almost all definitions include, at minimum, FFP (Resnik 2019; Resnik et al. 2015; Resnik, Peddada, and Brunson 2009).”
Self-citationAI hallucinations can’t be stopped — but these techniques can limit their damage
Jones, N. (2025) · Nature · doi:10.1038/d41586-025-00068-5 · Cited by 117 · cited 1×
“The problem of GenAI hallucinated citations is unlikely to go away completely because it is inextricably linked to how LLMs operate (Asai et al. 2026; Jones 2025; Kalai et al. 2025; Open AI 2025).”
From Baltimore to Bell Labs: Reflections on Two Decades of Debate about Scientific Misconduct
Resnik, D. B. (2003) · Accountability in Research · doi:10.1080/08989620300508 · Cited by 94 · cited 2×
“When scientists and policymakers crafted the U.S. federal government’s definition of research misconduct over two decades ago (Resnik 2003, 2019), they probably were not thinking about GenAI hallucinated citations.”
Self-citationIs it time to revise the definition of research misconduct?
Resnik, D. B. (2019) · Accountability in Research · doi:10.1080/08989621.2019.1570156 · Cited by 63 · cited 10×
“The key idea here is that a person’s mental state crosses the line from carelessness to recklessness when they demonstrate indifference to the harmful outcomes (such as FFP) of their behavior (Caron et al. 2025).”
InfluentialSelf-citationWriting a Scientific Review Article: Comprehensive Insights for Beginners
Amobonye, A., Lalung, J., Mheta, G., & Pillai, S. (2024) · The Scientific World JOURNAL · doi:10.1155/2024/7822269 · Cited by 55 · cited 1×
“The methodology involves identifying, classifying, and curating various references and extracting qualitative or quantitative data from them (Amobonye et al. 2024; Ghosh and Choudhury 2025).”
Research works printed but not linked: 4
- All European Academies. 2023. “European Code of Conduct for Research Integrity 2023 Revision.” Accessed December 24 2025. https://allea.org/wp-content/uploads/2023/06/European-Code-of-Conduct-Revised-Edition-2023.pdf.
- Office of Science and Technology Policy. 2000. “Research Misconduct Policy.” Federal Register 65 (235): 76260–76264. Accessed December 24, 2025. https://ori.hhs.gov/node/1465/printable/print.
- Retraction Watch. 2025. “The Case of the Fake References in an Ethics Journal.” December 2. Accessed December 31 2025. https://retractionwatch.com/2025/12/02/fake-references-chatgpt-journal-academic-ethics-springer-nature-whistleblowing/.
- Retraction Watch. 2026. “Librarian Finds ‘Preposterous Number’ of Fake References in Paper from Springer Nature Journal. March 6. Accessed March 7 2026. https://retractionwatch.com/2026/03/06/librarian-finds-preposterous-number-of-fake-references-in-paper-from-springer-nature-journal/.
Cited by
Hallucinated citations are polluting the scientific literature. What can be done?
Naddaf, M., & Quill, E. (2026) · Nature · doi:10.1038/d41586-026-00969-z · Cited by 36
Updated Guidance for Author Use of AI in Medical Publication
Flanagin, A., Perlis, R. H., & Bibbins-Domingo, K. B. (2026) · JAMA · doi:10.1001/jama.2026.16613 · Cited by 22
Moons, P. (2026) · European Journal of Cardiovascular Nursing · doi:10.1093/eurjcn/zvag148 · Cited by 2
Girdwood, T., & Manikonda, L. (2026) · Frontiers in Research Metrics and Analytics · doi:10.3389/frma.2026.1863790 · Cited by 3
Chen, C. K., Liang, S.-H., & Wang, T. (2026) · Scientometrics · doi:10.1007/s11192-026-05727-6 · Cited by 1
“The emerging consensus on hallucinated citations as potential research misconduct was articulated by Resnik and Hosseini (2026).”
Metadata
- Authors
- Resnik, D. B., & Hosseini, M.
- Affiliations
- National Institutes of Health (US)
- Northwestern University (US)
- Corresponding author
- Hosseini, Mohammad
- Venue
- Accountability in Research
- Publisher
- Informa UK Limited
- Published
- 2026-03-15
- Volume
- 33
- Issue
- 7
- Pages
- 2645390-2645390
- ISSN
- 0898-9621 (ISSN-L)
- 1545-5815
- PMC
- PMC13051339
- OpenAlex
- W7136066094
- Semantic Scholar
- cafd28af80633d644bc19381c06513e41922c91f
- Cited by
- 26
- References
- 27
- Citations by year
- 2026: 26
- FWCI
- 146.29
- Type
- journal article
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- creativecommons.org/licenses/by/4.0/ (vor)
- doi.org/10.1080/policies-text-data-mining (tdm)
- Locations
- Accountability in Research (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by)
- Funders
- National Center for Advancing Translational Sciences (UM1TR005121)
- National Institutes of Health (Intramural Research Program)
- Topics
- Social Sciences › Education › Academic integrity and plagiarism
- Physical Sciences › Computer Science › Artificial Intelligence › Ethics and Social Impacts of AI
- Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
- Keywords
- generative AI, research misconduct, AI hallucination, citation verification, citation practices, bibliometric analysis
- MeSH terms
- Generative Artificial Intelligence, Humans, United States, Scientific Misconduct
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Hallucinated citations produced by generative artificial intelligence may constitute research misconduct when citations function as data in scholarly papers. https://app.cochat.ai/papers/10.1080/08989621.2026.2645390
- The paper
- Resnik, D. B., & Hosseini, M. (2026). Hallucinated citations produced by generative artificial intelligence may constitute research misconduct when citations function as data in scholarly papers. Accountability in Research. https://doi.org/10.1080/08989621.2026.2645390
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.