Bringing Sanity to the Scientific Record
Read the preprint · MetaArXiv, 27 Oct 2026 · doi:10.31222/osf.io/uwqxa_v1
40.7% of citations to retracted papers still treat them as support.
51,626 citations judged by three AI models · Phase 2 census · pending human review
Forensic metascience for citations. We track retracted research that keeps spreading through the published literature, and publish every result free.
Retracted, and still built on.
The retracted papers at the top of our Lists, after their retraction. This measures reach, not harm.
papers published after the retraction cite one of 10 retracted papers as support
of those papers acknowledge grants
of those are systematic reviews or meta-analyses (13) or trials (13)
trials registered after the retraction list one among their references, enrolling or planning 1,669 people
Where the retracted work went.
Each story says how far the chain provably goes, and no further. Every fact names its source.
Germany’s sepsis guideline repeats a finding that depends on a retracted trial.
Take the retracted trial out of the review behind the guideline, and the finding is no longer statistically significant.
Read the story → Caught, two years lateA review counted a trial that its own reference list labels “Retracted”, and was retracted itself.
A 2024 review of sildenafil in threatened preterm labour. Its own rules excluded retracted articles.
Read the story → Sixteen years onThe Tai chi trial that would not stay retracted.
Retracted in 2010. Still counted in reviews in 2024, 2025 and 2026.
Read the story → Cited by studies that enrol peopleStudies enrolled patients while citing papers retracted after a misconduct investigation.
597 people enrolled so far, in studies that list retracted nerve-block papers among their references.
Read the story → The bigger picture93 trials failed a trust check. Journals had flagged 3.
Retraction marks about 1 in 500 randomised trials. Independent checks find far more that could not be confirmed as trustworthy.
Read the story →Who keeps citing retracted work?
Everyone named was written to before publication, with the evidence. Their replies are printed beside their entries.
The retracted papers most often still cited as support.
What each one claimed, why its retraction matters, and what it reached after it was retracted: later papers, reviews, trials and the grants behind them.
See the list → JournalsThe journals whose papers most often cite retracted work as support.
Ranked two ways: by count, and by share among journals with 100 or more judged citations.
See the list → AuthorsThe authors whose papers most often cite retracted work as support.
Every support citation in each listed author’s papers was read again before publication.
See the list →Retracted papers keep getting cited.
Hardly anyone reads a retraction notice. In one large study of biomedicine, only 5.4% of sentences citing a paper after its retraction said it had been retracted.1
of retracted papers were cited again after the retraction: 32,800 of 51,558.
citations, in all, to papers that had already been retracted.
retracted papers were still being cited in 2025 or 2026.
Source: Retraction Watch database via Crossref, cross-checked with the OpenAlex citation graph. Updated 6 Nov 2026, 11:45 UTC. What these numbers don’t say →
Retracted2 Feb 2010
Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children
The paper behind the MMR vaccine scare.
OpenAlex yearly counts start in 2012. *2026 is year to date.
This one is famous, so most papers warn about it. We read the 228 citations in open-access papers: 177 flag it as retracted or unreliable, and 19 still cite it as support. Across all retracted papers, only 3.3% of citations carry any warning.
Show the numbers
| Year | Citations |
|---|
Then three AI models read 55,688 citations to retracted biomedical papers.
Every citation made after retraction to a PubMed-indexed paper, wherever the citing paper is open access. On a blind sample the models matched a separate AI reader 94.7% of the time. These figures are pending human review.
40.7% of citations to retracted papers still treat them as support.
Out of every 100 citations made after a paper was retracted, 41 treat it as support, 56 mention it in passing, and just 3 flag it as retracted or unreliable.
Out of every 100 citations:
- 41treated it as support40.7% · 21,037 of 51,626
- 56mentioned it in passing55.9% · 28,863 of 51,626
- 3flagged it as retracted or unreliable3.3% · 1,726 of 51,626
Why it matters. A retraction is the journal telling readers not to rely on a paper. When later papers still cite it as support, its claims keep spreading into new research, reviews and the advice doctors read.
Source: SaneCite Phase 2 census: 51,626 citations to 12,328 retracted papers, from open-access papers citing retracted biomedical papers (PubMed-indexed, retracted 1990–2023). Read by three AI models; pending human review. Data as of 14 Oct 2026.
Source: SaneCite Phase 2 census, 51,626 citations, data as of 14 Oct 2026 · sanecite.com/charts
Show the numbers
| Label | Citations | Share (95% CI, %) |
|---|---|---|
| Treated it as support | 21,037 | 40.7% (39.9–41.5) |
| Mentioned it in passing | 28,863 | 55.9% (55.1–56.8) |
| Flagged it as retracted or unreliable | 1,726 | 3.3% (2.6–4.3) |
| All judged | 51,626 | 100% |
51,626 citations to 12,328 retracted papers, each made in a year after the retraction. Left out: 4,044 citations to paperwork retractions (no fault found in the science) and 18 the three readers could not decide. Not corrected for reader error. Corrected using one AI reader's checks of 150 citations, the estimate is 38.1% (95% CI 33.5–41.8%). 95% confidence intervals (the likely range) come from re-running the numbers 400 times, drawing whole retracted papers at random (a cluster bootstrap).
Even when a paper’s published reference list showed the retraction, 2 in 5 citations still treated it as support (40.4% of 3,628).
Citations made more than ten years after a retraction still treat the paper as support 34.9% of the time. Shares are of the 51,626 citations the models could judge, after leaving out paperwork retractions. All charts, method and data → · Search every citation → · Charts, free to share →
A machine reads it all. You can check the machine.
2Open, so you don’t have to trust us.
The method is published before each run. The census code is public under the MIT license, with its data, on GitHub, so anyone can re-run it.
A flag is not a finding. Authors often couldn’t know: the notice never reached them. We measure that failure of the system, not of the scientists, and anyone named gets to reply.
Method and limits → · Our rules → · How accurate the reading is → · Every check, in the order it ran →
Already run on real papers. We have read and checked 68,022 citations so far: 63,026 citations of retracted papers, including the Phase 2 census (55,676 citations after its 20 Oct 2026 correction; 55,688 as first published), and 4,996 ordinary citations in our accuracy tests. Every census call is published with its passage, for anyone to check.
Pay for the human check.
The machine has read 55,688 citations. A check by people, designed and registered in advance, will confirm or correct the headline. Gifts pay for the readers and the computing, and once SaneCite has a nonprofit fiscal sponsor, every gift and every spend is published. The machine reading of the whole census cost $32.05, so the checking is cheap; what would fix the problem →
- Hsiao T-K, Schneider J. Continued use of retracted papers: temporal trends in citations and (lack of) awareness of retractions shown in citation contexts in biomedicine. Quantitative Science Studies 2021. 722 of 13,252 post-retraction citation contexts in PubMed Central acknowledged the retraction.
- SaneCite Campaign One, Phase 1: first run 13 Oct 2026; corrected count 16 Oct 2026 (paperwork retractions left out). Retraction Watch database via Crossref (68,211 retraction records, 6 Nov 2026). Citations from OpenAlex counts_by_year; a citation counts as post-retraction if it falls in a calendar year after the retraction year. Exhibit: OpenAlex W2117847125, 2012–2026.