The most recent number Retraction Watch has published for its own database is 63,000 retraction records, as of December 30, 2025. That database, hosted as the Crossref Retraction Watch dataset, is rebuilt most working days, so the true count today is higher. We pulled the dataset's own build marker on July 13, 2026, and it read July 10, 2026. Every figure below carries the date it was published, not a guess at this minute's count.
The closest thing to a full count is the Crossref Retraction Watch dataset, a public CSV that Retraction Watch keeps building and Crossref hosts and updates. Retraction Watch's own most recent published tally, in its year-end post dated December 30, 2025, put the database at 63,000 retraction records. Because the underlying file is rebuilt on most working days, the live number today is very likely above 63,000. We looked for a way to read today's exact row count ourselves and could not confirm one, so we report the last number Retraction Watch itself published, dated, rather than estimate today's figure. See what a retraction notice actually means if that term is new to you.
Retraction Watch's own year-end counts, each dated, trace the growth directly: 18,500 records at the database's October 2018 launch, 25,000 by April 2021, more than 37,000 at the end of 2022 (that year alone added over 4,600), 45,000 at the end of 2023, and 63,000 at the end of 2025. Separately, Nature's own analysis, reported by Richard Van Noorden in December 2023, found that journals retracted more than 10,000 papers in 2023 alone, a new annual record, and that the retraction rate (retractions as a share of papers published that year) had more than trebled over the prior decade. Retraction Watch's own 2022 review put the rate at roughly 8 retractions per 10,000 papers published that year.
The growth is not even across publishers. A February 2026 analysis of 46,087 retractions across ten major publishers, using the Retraction Watch database and covering 1997 to 2026, found rates ranging from 3.97 retractions per 10,000 published articles at Elsevier to 320.02 per 10,000 at Hindawi, with Springer Nature at 9.06, PLOS at 26.82, and IOS Press at 283.77. Retraction Watch reported the full breakdown on May 11, 2026. A high rate at one publisher is not proof of worse conduct there; it can also reflect how aggressively that publisher looks for problems and issues notices once found.
For biomedical literature specifically, an analysis of Retraction Watch Database records from 2014 to 2023 found a median of 390 days from publication to retraction (mean 580 days), and found that gap shrinking by about 89 days a year over the decade studied, meaning journals have gradually gotten faster at retracting. That study downloaded its data on January 15, 2024. A separate, longer analysis spanning 1975 to 2024 (data cutoff December 30, 2024) put the overall median time to retraction for medical publications at 562 days. Neither study covers every research field, so treat both as biomedical-specific, not universal, figures.
Yes, and this is the part the headline retraction count does not show. A study in Quantitative Science Studies examined 7,813 retracted papers indexed in PubMed, 169,434 citations to them, and 48,134 citation contexts drawn from full text. About 27.5 percent of those citation contexts occurred after the paper's retraction date, and only 5.4 percent of the post-retraction citations acknowledged that a retraction had happened. A separate matched-control study tracking citation frequency before and after retraction found average yearly citations to retracted papers fall from about 4.85 to 1.93 per year, a real drop, but not to zero, while matched non-retracted papers kept climbing over the same period.
Every figure above traces back to the Crossref Retraction Watch dataset or to a study built on it. The dataset's own README, which we checked live on July 13, 2026, lists 20 fields: Record ID, Title, Subject, Institution, Journal, Publisher, Country, Author, URLS, ArticleType, RetractionDate, RetractionDOI, RetractionPubMedID, OriginalPaperDate, OriginalPaperDOI, OriginalPaperPubMedID, RetractionNature, Reason, Paywalled, and Notes. It is maintained by Retraction Watch and hosted publicly by Crossref, which acquired the database from The Center For Scientific Integrity under an agreement the two organizations announced on September 12, 2023. Retraction Watch updates it most working days. The version we checked on July 13, 2026 carried a build date of July 10, 2026. If you pull the file yourself later, expect a higher cumulative count than the 63,000 cited here: that is the point of dating every number instead of stating one as final.
We run the same Crossref retraction check, DOI by DOI, on every paper before it reaches an issue of the newsletter. If a paper's DOI carries a retraction notice, we do not run it, full stop. You can run the identical check yourself on any paper with Lucent's free DOI checker, and see the fuller picture of what a single check can and cannot tell you at how to tell if a study is trustworthy.
Every two days: brand-new papers in plain language, each screened for retractions and a live DOI. Three free every issue.
What is a paper retraction · How to tell if a study is trustworthy · Check a DOI · Latest issue · Privacy