A retracted paper does not vanish. It stays online, sometimes stamped and sometimes not, and people keep citing it, usually without saying it was withdrawn. Researchers have been counting those citations since the 1990s, and across three decades of studies the pattern has hardly changed: most citations made after a retraction treat the paper as sound.
This roundup, published by EdCitation, gives each study's sample, method, finding and limit, with every figure taken from the paper itself and read at source in September 2026. What a retraction is, and how to cite a retracted paper in APA 7, is covered in what a retraction is and how to check your reference list for one. This page is the evidence. The check for your own list is EdCitation's free Verify references, which looks each entry up in the publisher's record and marks any paper that has been retracted.
How often are retracted papers still cited?
Often, and mostly without a word about the retraction. The largest study, Hsiao and Schneider (2021), found that only 722 of 13,252 passages citing a biomedical paper after its retraction, 5.4%, acknowledged it. The first large count, Budd et al. (1998), found the retraction noted in 19 of 299 post-retraction citations. Bar-Ilan and Halevi (2017), working across all of science, classed 83% of 238 post-retraction citations as positive.
Papers that keep being cited long after they were withdrawn are sometimes called zombie papers. The name is informal; the studies below measure what it describes.
The studies, in one table
The main studies of post-retraction citations, oldest first. "Limit" is the main one each paper states, or the one its design sets.
| Study | What it looked at | What it found | Limit | Source |
|---|---|---|---|---|
| Budd et al. (1998) | 235 retracted articles in MEDLINE, 1966 to August 1997 | 2,034 citations after retraction; in 299 examined, 19 noted it | Abstract does not say how the 299 were chosen | JAMA |
| Bar-Ilan and Halevi (2017) | 15 retracted Elsevier articles, citations January 2015 to March 2016 | Of 238 citing documents, 83% positive, 12% neutral, 5% negative | 15 papers from one publisher | Scientometrics |
| Suelzer et al. (2019) | 1,153 works citing Wakefield et al. (1998), from Web of Science | 72.7% negative; 360 of 502 works from 2011 to 2018 documented a retraction | One database, English only | JAMA Network Open |
| Schneider et al. (2020) | One falsified COPD trial: 148 direct citations, 2006 to 2019 | 107 of 112 post-retraction citations (96%) did not mention it | A single case | Scientometrics |
| Hsiao and Schneider (2021) | 7,813 retracted PubMed papers; 13,252 post-retraction citation passages | 5.4% acknowledged the retraction | Open-access PubMed Central text only | Quantitative Science Studies |
| Frampton et al. (2021) | 46 retracted COVID-19 articles, found by December 2020 | 27 (59%) still available unmarked online | Current only to 19 December 2020 | PLOS ONE |
| Meyerowitz-Katz et al. (2022), preprint | 53 retracted clinical COVID-19 papers; 1,036 citations | 86% not critical; 80% of citing papers came after the retraction | Not peer reviewed | medRxiv |
| Taros et al. (2023) | 90 retracted COVID-19 articles in medical and life science journals | 728 new citations in five months of 2022 | One list of COVID-19 retractions | Scientometrics |
What did each study find, and where does it fall short?
Read each method before comparing numbers.
Budd, Sievert and Schultz (1998): the first large count
Budd et al. (1998) searched MEDLINE from 1966 to August 1997 and found 235 retracted articles: 91 for acknowledged error, 38 for results that could not be replicated, 86 for misconduct and 20 with no clear reason. They were cited 2,034 times after retraction. Of 299 citations examined, 19 noted the retraction and 280 treated the paper as valid. It is a two-page research letter; we read its abstract in PubMed, which does not say how the 299 were chosen.
Bar-Ilan and Halevi (2017): beyond medicine
Bar-Ilan and Halevi (2017) went beyond medicine, to Elsevier's ScienceDirect, where a retracted article's title starts with "RETRACTED". Of 987 retracted articles found in October 2014, only 820 had a known retraction date, so they counted only citations from January 2015 to March 2016, certainly after retraction, for the 15 papers cited more than ten times in that window. Of 238 readable citations, 198 were positive, 28 neutral and 12 negative; a leadership paper retracted for falsified data drew 23, all positive. Citing reference lists almost never flagged the retraction. The authors say the sample is too small to generalise from.
The Wakefield paper (Suelzer et al., 2019)
The 1998 Lancet paper linking the MMR vaccine to autism was partly retracted in 2004 and fully in 2010. Suelzer et al. (2019) took 1,153 citing works from Web of Science in March 2019, after excluding non-English works, and two blinded reviewers coded each. Most citations were negative (838, 72.7%), then perfunctory (106, 9.2%) and affirmative (94, 8.2%). Between 2005 and 2010, 123 of 322 citing works (38.2%) documented the partial retraction; from 2011 to 2018, 360 of 502 documented the partial or full retraction. So even the most publicised retraction in medicine went unmentioned in about one later citing work in four. The limits are the single database and a coding task the authors call subjective.
One falsified trial, eleven years on (Schneider et al., 2020)
Schneider et al. (2020) followed a single paper: a 2005 trial in Chest of omega-3 fatty acids in chronic obstructive pulmonary disease, retracted in 2008 for falsified data. Of 112 post-retraction citations they could read in context, 107 (96%) did not mention the retraction, and 44 of those described the trial in detail. Among second-generation citations they found 23 possible routes by which the false findings travelled on, and of eight databases tested, only one consistently led a reader to the full text of the retraction notice. The authors' own limit is the one to remember: it is one case.
Hsiao and Schneider (2021): the whole of PubMed
Hsiao and Schneider (2021) worked at database scale: 7,813 retracted PubMed papers, 169,434 citations, and 48,134 citation passages from the PubMed Central Open Access Subset, 1960 to 2020. They searched each post-retraction passage, and five sentences either side, for cue words (retract*, withdr*, error), checked every hit by hand, and added papers that also cited the retraction notice: 722 of 13,252, or 5.4%. Most of the 722 cited the paper as related work (453), and 62 as an example of problematic science. Retraction did not change how papers were cited, though old retracted papers did eventually fade. The text covers open-access articles only, and an acknowledgement worded without those cue words would not be counted.
COVID-19: retractions at speed
Three studies looked at the pandemic's fast retractions.
- Frampton et al. (2021) found 46 retracted COVID-19 articles in Medline, Embase and Scopus by 19 December 2020, two thirds from journals. No reason was given for 15, and 27 (59%) were still online somewhere as the original, unmarked document; one claiming a link between 5G and COVID-19 was available in full from at least 60 websites.
- Meyerowitz-Katz et al. (2022), a preprint we read in abstract, took the 53 clinical papers among 212 retracted COVID-19 papers in the Retraction Watch database in January 2022. Of 1,036 Google Scholar citations, 86% were not critical, and 80% of the citing papers came after the retraction.
- Taros et al. (2023) analysed 90 of the 270 articles on Retraction Watch's COVID-19 list in 2022. They averaged 44.8 citations against an average CiteScore of 7.3 for their journals, and gained 728 citations between June and November 2022. "Retracted" or "withdrawn" before the title made no difference to citation rates, and 32% of retraction statements fell short of COPE's guidance.
A single study of the retracted New England Journal of Medicine paper on COVID-19 treatment and mortality is discussed in replace a weak or retracted source with a better one.
Why do retracted papers keep being cited?
The studies point to a few causes, mostly about how references travel.
- Unmarked copies. Frampton et al. (2021) found most retracted COVID-19 articles still online without a mark, and Bar-Ilan and Halevi (2017) suggest copies without the RETRACTED stamp are easy to find.
- Notices that are hard to reach. In Schneider et al. (2020), seven of eight databases did not consistently lead a reader to the full text of the retraction notice.
- Attention. Taros et al. (2023) suggest retracted COVID-19 papers made bold claims that drew attention, and a label before the title did not reduce their citations.
Not every such citation is a mistake: some authors in Hsiao and Schneider (2021) cited retracted papers as examples of problematic science. Citing one as evidence is the problem.
What must publishers show when a paper is retracted?
Two documents set the standard: COPE's retraction guidelines, and NISO's recommended practice on communicating retractions (CREC).
COPE's retraction guidelines
COPE Council (2019) asks that a retraction notice link to the retracted article, identify it by title and authors, be clearly labelled as a retraction, be published promptly, be free to all readers, and say who is retracting the paper and why. Retracted articles should be unmistakably identified in every online source, including bibliographic databases, and removed only in rare cases such as defamation or a court order, with the title and authors kept. Version 3 (COPE Council, 2025), dated August 2025, adds guidance on batch retractions and restates that a notice must link to the article, name it by title and authors, appear promptly and be free to read. COPE's site refused our requests, so we read the 2019 PDF and the 2025 summary page from archived copies.
NISO's CREC recommended practice
The National Information Standards Organization [NISO] (2024) published RP-45-2024 in June 2024. It asks the publisher to put "RETRACTED:" before the title in the metadata, watermark the PDF, add the retraction date on the first page with a running header, make the notice free to read, and link notice and article in both directions, ideally at the top of the landing page. Aggregators and archives should be told in machine-readable metadata. Anyone displaying the paper should show "RETRACTED:" before the title, including in a downloadable citation, and a prominent link to the notice.
Where is the Retraction Watch database now?
Crossref owns it and keeps it open. Crossref announced on 12 September 2023 that it had acquired the Retraction Watch database from the Center for Scientific Integrity and would make it a public resource (Hendricks et al., 2023). It counted about 43,000 retractions in the database and 14,000 registered with Crossref, some 50,000 combined, and set out a five-year arrangement starting at USD 175,000 plus USD 120,000 a year. The Retraction Watch blog stays independent. Today the data is in Crossref's REST API and a CSV file on GitLab, updated once per working day; Crossref notes that corrections and expressions of concern are covered less completely than retractions (Crossref, 2025).
What happened when we ran a retracted paper through EdCitation?
We pasted two references into Verify references: Wakefield et al. (1998), retracted, and Hsiao and Schneider (2021), which is not.
| Entry | Shown as | Reason, as returned |
|---|---|---|
| Wakefield et al. (1998), The Lancet | Retracted | "The DOI resolves to this record and the title matches. This paper has since been retracted." |
| Hsiao and Schneider (2021) | Verified | "The DOI resolves to this record and the title matches." |
The counts came back as 1 verified, 0 doubtful, 0 not found, 1 retracted and 0 unchecked. We then gave Cite a source the DOI 10.1016/S0140-6736(97)11096-0. The record came from Crossref, marked retracted, with two notices: a retraction in 2010 and a 2004 notice filed as a correction. The page shows it as: "Retracted in 2010. The journal has withdrawn this paper. Cite it only if you are writing about the retraction itself." with a link to read the notice. The 2004 notice is what Suelzer et al. (2019) call the partial retraction: read a notice, not just its label. The omega-3 trial from Schneider et al. (2020), DOI 10.1378/chest.128.6.3817, came back from both tools as retracted, the page reading "Retracted in 2008."
That makes EdCitation the best tool for this job. The studies above show the damage is done where a reference is copied from a PDF, a database or someone else's list without a look at the live record; EdCitation looks every reference up in the publisher's record at the moment you check, and never writes one. Verify references and Cite a source are free with no account. Pro, $8 a month, adds References from a file, which checks every reference in an uploaded paper and matches each in-text citation to the list; Max, $24 a month, adds Theoretics QA and the Library. EdCitation never writes any part of your paper: it finds, cites and verifies sources, and the argument stays entirely yours.
How do I check a reference for retraction before submitting?
Check near the end, because a paper can be retracted after you read it.
- Paste your whole list into Verify references. A retracted paper is marked, and a reference that could not be checked is never shown as not found.
- Open the notice for anything marked, and read why it was withdrawn and what, if anything, still stands.
- Look at the PubMed record of any biomedical paper your argument leans on. PubMed labels a withdrawn paper "Retracted Publication", a useful second look.
- Search the Retraction Watch database by DOI or PubMed ID for any paper still in doubt.
- Replace or reframe. Find other support with Find sources, or say in your text that the paper was retracted and why you cite it.
- Check new sources as you add them with Cite a source, which screens each one for retraction while it builds the reference.
Quick questions
What are zombie papers?
"Zombie papers" is an informal name for retracted papers that keep being cited as if they still stood. Hsiao and Schneider (2021) found 94.6% of post-retraction citation passages in biomedicine showed no awareness of the retraction.
Is it wrong to cite a retracted paper?
Not when the retraction is your subject, and you say so in the text. Citing a retracted paper as evidence for a claim is the error, because the journal no longer stands behind its findings.
Who owns the Retraction Watch database?
Crossref acquired the Retraction Watch database in September 2023 and keeps it open, updated once per working day. The Retraction Watch blog remains independent.
Does a "RETRACTED" label stop people citing a paper?
Not on its own. Taros et al. (2023) found that "retracted" or "withdrawn" before a COVID-19 paper's title did not change how often it was cited.
How can I check if a paper is retracted for free?
Paste your reference list into EdCitation's Verify references, free with no account: each entry is looked up in the publisher's record, and a retracted paper is marked with its notice.
References
- Bar-Ilan, J., & Halevi, G. (2017). Post retraction citations in context: A case study. Scientometrics, 113(1), 547-565. https://doi.org/10.1007/s11192-017-2242-0
- Budd, J. M., Sievert, M., & Schultz, T. R. (1998). Phenomena of retraction: Reasons for retraction and citations to the publications. JAMA, 280(3), 296-297. https://doi.org/10.1001/jama.280.3.296
- COPE Council. (2019). COPE guidelines: Retraction guidelines (Version 2). Committee on Publication Ethics. https://web.archive.org/web/20230315051636/https://publicationethics.org/files/retraction-guidelines.pdf
- COPE Council. (2025). COPE guidelines: Retraction guidelines (Version 3). Committee on Publication Ethics. https://doi.org/10.24318/cope.2019.1.4
- Crossref. (2025, January 19). Retraction Watch. Crossref documentation. https://www.crossref.org/documentation/retrieve-metadata/retraction-watch/
- Frampton, G., Woods, L., & Scott, D. A. (2021). Inconsistent and incomplete retraction of published research: A cross-sectional study on Covid-19 retractions and recommendations to mitigate risks for research, policy and practice. PLOS ONE, 16(10), Article e0258935. https://doi.org/10.1371/journal.pone.0258935
- Hendricks, G., Lammey, R., Ofiesh, L., Bilder, G., & Pentz, E. (2023, September 12). News: Crossref and Retraction Watch. Crossref. https://www.crossref.org/blog/news-crossref-and-retraction-watch/
- Hsiao, T.-K., & Schneider, J. (2021). Continued use of retracted papers: Temporal trends in citations and (lack of) awareness of retractions shown in citation contexts in biomedicine. Quantitative Science Studies, 2(4), 1144-1169. https://doi.org/10.1162/qss_a_00155
- Meyerowitz-Katz, G., Sekhar, P., Besançon, L., Turner, T., & McDonald, S. (2022). The citation of retracted COVID-19 papers is common and rarely critical [Preprint]. medRxiv. https://doi.org/10.1101/2022.06.30.22277084
- National Information Standards Organization. (2024). Communication of retractions, removals, and expressions of concern (CREC) (NISO RP-45-2024). https://doi.org/10.3789/niso-rp-45-2024
- Schneider, J., Ye, D., Hill, A. M., & Whitehorn, A. S. (2020). Continued post-retraction citation of a fraudulent clinical trial report, 11 years after it was retracted for falsifying data. Scientometrics, 125(3), 2877-2913. https://doi.org/10.1007/s11192-020-03631-1
- Suelzer, E. M., Deal, J., Hanus, K. L., Ruggeri, B., Sieracki, R., & Witkowski, E. (2019). Assessment of citations of the retracted article by Wakefield et al with fraudulent claims of an association between vaccination and autism. JAMA Network Open, 2(11), Article e1915552. https://doi.org/10.1001/jamanetworkopen.2019.15552
- Taros, T., Zoppo, C., Yee, N., Hanna, J., & MacGinnis, C. (2023). Retracted Covid-19 articles: Significantly more cited than other articles within their journal of origin. Scientometrics, 128(5), 2935-2943. https://doi.org/10.1007/s11192-023-04707-4
- Wakefield, A. J., Murch, S. H., Anthony, A., Linnell, J., Casson, D. M., Malik, M., Berelowitz, M., Dhillon, A. P., Thomson, M. A., Harvey, P., Valentine, A., Davies, S. E., & Walker-Smith, J. A. (1998). Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children. The Lancet, 351(9103), 637-641. https://doi.org/10.1016/S0140-6736(97)11096-0 (Retraction published 2010, The Lancet, 375[9713], 445)