Contents
Academy · Research Publication · Fraud Detection & Retraction Systems
The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction
A systematic screen of 20,621 biomedical papers found inappropriate image duplication, a strong marker of fabrication, in approximately one in twenty-five. The discovery that a widely cited Alzheimer's paper rested on manipulated images indicates that fraud has reached even the most consequential areas of the literature.
A fake or fraudulent paper is one whose central data, images, or results have been fabricated, falsified, or manipulated to misrepresent what was actually done or found. Detecting it requires a deliberate combination of content-based checks (examining the data, images, and methods themselves) and non-content-based checks (examining the surrounding submission behavior, writing style, and review history), because fabrication is specifically designed to pass a casual reading. Once detected, the available response is retraction, which only works to protect the literature if it is executed transparently and is recorded in a way later readers will actually encounter.
Learning objectives
- The measured prevalence of image duplication found by a large systematic screen of biomedical papers
- The distinct content-based and non-content-based methods used to detect fabricated papers
- Why transparent retraction notices matter as much as detection itself
Common misconceptions
- “Fraudulent papers are mostly a problem of obscure, low-tier journals.” The manipulated-image findings behind a leading amyloid-beta Alzheimer's disease paper, reported by Science journalist Charles Piller in 2022, and the series of retracted room-temperature superconductor claims from a research group whose findings were initially published in high-profile venues, both demonstrate that fabricated or manipulated data has reached journals and research programs with substantial downstream influence, not only marginal ones.
Prevalence of image manipulation
Elisabeth Bik, Arturo Casadevall, and Ferric Fang's 2016 study in mBio visually screened 20,621 papers published across 40 biomedical journals between 1995 and 2014 and found problematic figures, consistent with inappropriate duplication or manipulation, in 3.8% of papers, with about half of those showing features suggestive of deliberate manipulation rather than innocent error1. That rate, applied across the volume of biomedical literature published annually, implies a substantial absolute number of affected papers, and the study's methodology, pure visual inspection without specialized forensic software, likely understates the true prevalence, since more sophisticated manipulation would not necessarily be visible to the human eye1. The consequences of undetected fabrication in influential work were illustrated by Charles Piller's 2022 Science investigation, which reported that a foundational, widely cited paper on amyloid-beta oligomers in Alzheimer's disease appeared to contain manipulated images, a finding that raised questions about years of subsequent research built on that specific result6.
Content-based detection of data and image fabrication
Content-based detection methods examine a paper's substance directly. This includes image-forensics software that flags duplicated, rotated, or spliced figure elements, of the kind used in the Bik, Casadevall, and Fang screening study1, requests for raw underlying data that a legitimate researcher should be able to produce on demand, and, in genomics-adjacent fields, checks of nucleotide sequences against public repositories to confirm they correspond to genuine biological material rather than fabricated entries. Guillaume Cabanac and Cyril Labbé's research on algorithmically generated papers extended content-based detection into text itself, identifying papers produced by text-generation software designed to mimic the surface structure of a scientific article without genuine underlying research2. Their follow-up work on "tortured phrases," unusual, garbled synonym substitutions such as replacing a standard technical term with an odd paraphrase, apparently introduced by automated paraphrasing tools used to evade plagiarism detection, gave editors and screening software a specific, searchable marker of likely fabricated or plagiarism-laundered text3.
Non-content detection based on submission behavior
Non-content detection looks at patterns surrounding a submission rather than the scientific content itself. Jana Christopher's analysis of what she terms the "dark arts" of image and data manipulation for journals and reviewers describes recurring behavioral red flags associated with fraudulent submissions originating from paper mills: templated manuscript structures reused across many superficially different papers, author lists that change between submission and acceptance, and reviewer reports that appear suspiciously fast, generic, or uncritical4. Jennifer Byrne and Jana Christopher's related work recommends that journals and reviewers specifically watch for these submission-pattern anomalies alongside content checks, because paper mills often operate at a scale where the same underlying template, methods section, or figure set reappears, with cosmetic variation, across dozens or hundreds of submissions to different journals4. Both content and non-content methods carry substantive limitations, including the time and specialized tooling they require and the documented risk that automated text-screening tools can wrongly flag legitimate writing by non-native English speakers as suspicious, a concern that argues for human review of any automated flag rather than automatic rejection.
Limitations of both detection approaches
Neither detection category is infallible, and overstating their reliability risks both false confidence and unfair treatment of honest authors. Elisabeth Bik's own follow-up commentary on image-screening work has noted that visual and even semi-automated image forensics can miss sophisticated manipulation specifically designed to evade detection, while also occasionally flagging legitimate images that happen to share superficial visual similarity for innocent reasons, such as a standard control image reused appropriately across a research group's related papers with proper disclosure1. Cabanac and Labbé's own writing on tortured-phrase detection is similarly explicit that the method flags a textual pattern correlated with likely fabrication or plagiarism-laundering, not a conclusive proof of either, and that editors should treat a tortured-phrase flag as grounds for closer manual review rather than an automatic rejection3. On the non-content side, Christopher's guidance for editors is explicit that behavioral red flags, templated structure, unusually fast reviews, are probabilistic indicators that should prompt investigation, not standalone grounds for retraction or rejection, since legitimate research groups publishing a genuine series of closely related studies can produce superficially similar patterns for entirely honest reasons4. The practical implication is the same for both detection categories: automated and pattern-based tools are best used to direct limited expert attention toward the submissions that most warrant closer examination, not to generate an automatic verdict, which is also why shared, cross-publisher detection infrastructure such as the STM Integrity Hub has been built specifically to pool expert review capacity rather than to replace it entirely with automated rejection7.
Transparent retraction as the necessary final step
Detecting fabrication only protects the literature if it leads to a retraction notice that later readers will find and understand. The Committee on Publication Ethics (COPE) retraction guidelines specify that notices should clearly state the reason for retraction and be freely accessible rather than vague or paywalled, precisely because an unclear or quietly issued retraction leaves a flawed paper functionally still in circulation for anyone who encounters it through search or citation without reading the notice carefully5. A joint COPE and STM research project specifically examining paper mills reinforced this point from the detection side, recommending that publishers document and share, within appropriate data-protection limits, the specific evidence behind a paper-mill-linked retraction so that other journals can recognize the same source if it resurfaces elsewhere8. Cristina Candal-Pedreira and colleagues' 2022 study tracing retracted papers back to specific paper mills found that transparent, well-documented retraction notices were essential to their own ability to trace and categorize the affected papers after the fact, illustrating concretely how a vague notice would have left even a dedicated research-integrity investigation unable to reconstruct what had actually happened9. The U.S. Office of Research Integrity's published case summaries of confirmed misconduct findings follow a comparable transparency principle at the institutional level, documenting the specific nature of confirmed fabrication or falsification in each case rather than issuing a vague finding of "misconduct" without detail, a practice that gives the broader research community enough specificity to recognize similar patterns in other work10. The documented case of retracted room-temperature superconductor claims, where specific papers were withdrawn by their own journals after independent labs failed to replicate the central result and data-integrity concerns emerged, illustrates both aspects of this principle: the retraction protected the field from continuing to build on an unreplicated result, but only because it was issued with enough specificity and visibility that the broader research community registered it quickly rather than continuing to cite the original claim. Taken together, detection and transparent retraction function as two halves of a single system rather than independent safeguards: detection without a clear, accessible notice leaves later readers unable to learn what was found, and a notice issued without rigorous prior detection risks being vague, contested, or simply wrong about what actually happened, so an editorial process that genuinely protects the literature needs to invest comparable care in both halves rather than treating either one as sufficient on its own. The resource implication is worth stating directly for smaller journals without a large in-house integrity team: joining shared infrastructure such as the STM Integrity Hub, and adopting COPE's model flowcharts rather than drafting bespoke internal procedures from scratch, lets a smaller editorial operation benefit from detection and process standards it would not have the capacity to build independently.
Key points
Key points
Detecting fabricated or manipulated papers requires both content-based checks (image forensics, raw-data requests, text analysis for tortured phrases or AI-generated text) and non-content checks (submission-pattern and reviewer-behavior red flags), and detection only protects the scientific record if it is followed by a clearly worded, freely accessible retraction notice rather than a quiet or vague correction.
Practical application
Practical application
- Request raw data and run image-forensics checks on any submission with figure-based results. This should be built into standard editorial workflow rather than reserved for cases that already look suspicious, since the Bik, Casadevall, and Fang study found meaningful duplication rates even in top-tier journals.
- Monitor submission-pattern warning signs as well as content-level indicators. Templated structures reused across papers, shifting author lists, and uncritical or unusually fast peer reviews are documented markers associated with paper-mill submissions.
- Where a retraction is being issued, follow COPE's notice requirements exactly. Clearly state the reason, link the notice to the original article, and ensure it is freely accessible so it reaches readers who encounter the paper through search or citation rather than the journal's own site.
Frequently asked questions
A large systematic visual screen of 20,621 papers found problematic figures in 3.8% of them, with roughly half showing features consistent with deliberate manipulation; because the screening relied on visual inspection without specialized forensic tools, this figure likely understates the true prevalence.
No. Documented cases include manipulated images underlying a foundational, highly cited Alzheimer's disease paper and retracted room-temperature superconductor claims originally published in high-profile venues, showing fabrication has reached influential research programs, not only marginal ones.
They are unusual, garbled synonym substitutions, such as an odd paraphrase replacing a standard technical term, that research has linked to automated paraphrasing tools used to evade plagiarism detection; they give editors and screening software a specific, searchable marker associated with fabricated or plagiarism-laundered text.
Sophisticated fabrication may not be visible even to careful visual inspection or standard plagiarism checks, which is why non-content detection, examining submission patterns, author-list changes, and reviewer behavior, adds a complementary layer that content checks alone can miss.
COPE guidelines specify that an effective notice clearly states the retraction's reason, is directly linked to the original article, and is freely accessible rather than vague or paywalled, since an unclear or hard-to-find notice leaves a flawed paper functionally still in circulation for later readers.
References
- Bik EM, Casadevall A, Fang FC. The Prevalence of Inappropriate Image Duplication in Biomedical Research Publications. mBio. 2016;7(3):e00809-16. read 10 October 2026. View on mBio
- Cabanac G, Labbé C. Prevalence of nonsensical algorithmically generated papers in the scientific literature. Journal of the Association for Information Science and Technology. 2021;72(12):1461-1476. read 10 October 2026. View on JASIST
- Cabanac G, Labbé C, Magazinov A. Tortured phrases: A dubious writing style emerging in science. arXiv:2107.06751. read 10 October 2026. View on arXiv
- Christopher J. The raw truth about paper mills. FEBS Letters. 2021;595(13):1751-1757. read 10 October 2026. View on FEBS Letters
- Committee on Publication Ethics (COPE). Retraction Guidelines. read 10 October 2026. View COPE's Retraction Guidelines
- Piller C. Blots on a field? Science. 2022;377(6604):358-363. read 10 October 2026. View on Science
- STM (International Association of Scientific, Technical and Medical Publishers). STM Integrity Hub. read 10 October 2026. View the STM Integrity Hub
- Committee on Publication Ethics and STM. Paper mills research project report. read 10 October 2026. View the COPE/STM paper mills report
- Candal-Pedreira C, Ross JS, Ruano-Ravina A, Egilman DS, Fernández E, Pérez-Ríos M. Retracted papers originating from paper mills: cross sectional study. BMJ. 2022;379:e071517. read 10 October 2026. View on BMJ
- U.S. Office of Research Integrity. Case Summaries. read 10 October 2026. View ORI Case Summaries
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 10). The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction. Pubrica. https://pubrica.com/academy/publications/fraud-in-scientific-publishing/
Pubrica Academy. “The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction.” Pubrica, 10 Oct. 2026, pubrica.com/academy/publications/fraud-in-scientific-publishing/.
Pubrica Academy (2026) The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction. Pubrica. Available at: https://pubrica.com/academy/publications/fraud-in-scientific-publishing/ (Accessed: 15 October 2026).
Pubrica Academy. “The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction.” Pubrica. October 10, 2026. https://pubrica.com/academy/publications/fraud-in-scientific-publishing/
Pubrica Academy, ‘The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction’ (Pubrica, 10 October 2026) <https://pubrica.com/academy/publications/fraud-in-scientific-publishing/> accessed 15 October 2026.
Pubrica Academy. The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction. [Internet]. Pubrica; 2026 Oct 10 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/publications/fraud-in-scientific-publishing/
Pubrica Academy, “The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction,” Pubrica, Oct. 10, 2026. [Online]. Available: https://pubrica.com/academy/publications/fraud-in-scientific-publishing/
@misc{/la_fake_papers_detection_prevention_retraction_2026,
author = {{Pubrica Academy}},
title = {The Impact of Fake Papers on Scientific Publishing: Detection, Prevention, and Retraction},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/publications/fraud-in-scientific-publishing/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Fraud Detection Screening Checklist
This checklist is intended for use during editorial or institutional review of a submission or published paper.
[ ] Ran image-forensics screening on all figures [ ] Requested raw underlying data and confirmed it was provided and plausible [ ] Screened text for tortured phrases or signs of algorithmically generated content [ ] Checked for submission-pattern red flags: templated structure, shifting author lists, unusually fast/uncritical reviews [ ] If fabrication is confirmed, drafted a retraction notice meeting COPE's clarity and accessibility requirements [ ] Cross-checked flagged non-native-English writing manually before treating automated flags as conclusive
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica's research integrity team screens manuscripts and datasets for fabrication indicators and helps editorial teams draft retraction notices that meet COPE's transparency standards.