Warnings fell from 9.4% of citations in 2009–2011 to 2.2% in 2024–2026.
The share of citations that flag the paper as retracted or unreliable, by the year the citing paper appeared.
Flagged it as retracted or unreliable, by years cited (number of citations)
Why it matters. A warning in the text is how the next reader learns a paper was withdrawn. As more papers cite retracted work, fewer of them pass that warning on.
Exploratory. Each year mixes papers retracted long ago and recently, so this shows the trend, not its cause.
Source: SaneCite Phase 2 census: 51,626 citations 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
| Year cited | Citations | Flagged as retracted | Cited as support |
|---|---|---|---|
| 1997 | 5 | 0 (0.0%) | 2 (40.0%) |
| 1998 | 6 | 0 (0.0%) | 3 (50.0%) |
| 1999 | 1 | 0 (0.0%) | 1 (100.0%) |
| 2000 | 6 | 1 (16.7%) | 3 (50.0%) |
| 2001 | 1 | 0 (0.0%) | 0 (0.0%) |
| 2002 | 1 | 0 (0.0%) | 1 (100.0%) |
| 2003 | 7 | 0 (0.0%) | 1 (14.3%) |
| 2004 | 11 | 1 (9.1%) | 6 (54.5%) |
| 2005 | 19 | 2 (10.5%) | 9 (47.4%) |
| 2006 | 33 | 3 (9.1%) | 10 (30.3%) |
| 2007 | 69 | 12 (17.4%) | 21 (30.4%) |
| 2008 | 75 | 5 (6.7%) | 24 (32.0%) |
| 2009 | 136 | 19 (14.0%) | 49 (36.0%) |
| 2010 | 205 | 17 (8.3%) | 77 (37.6%) |
| 2011 | 364 | 30 (8.2%) | 114 (31.3%) |
| 2012 | 644 | 65 (10.1%) | 227 (35.2%) |
| 2013 | 914 | 62 (6.8%) | 363 (39.7%) |
| 2014 | 1,102 | 73 (6.6%) | 439 (39.8%) |
| 2015 | 1,341 | 67 (5.0%) | 467 (34.8%) |
| 2016 | 1,598 | 52 (3.3%) | 596 (37.3%) |
| 2017 | 2,068 | 75 (3.6%) | 781 (37.8%) |
| 2018 | 2,021 | 79 (3.9%) | 772 (38.2%) |
| 2019 | 2,502 | 104 (4.2%) | 939 (37.5%) |
| 2020 | 3,415 | 106 (3.1%) | 1,330 (38.9%) |
| 2021 | 5,760 | 265 (4.6%) | 2,326 (40.4%) |
| 2022 | 7,173 | 193 (2.7%) | 3,236 (45.1%) |
| 2023 | 6,188 | 139 (2.2%) | 2,650 (42.8%) |
| 2024 | 8,570 | 160 (1.9%) | 3,698 (43.2%) |
| 2025 | 5,296 | 126 (2.4%) | 2,138 (40.4%) |
| 2026 | 2,095 | 70 (3.3%) | 754 (36.0%) |
The chart groups the years from 2009 in threes; earlier years have fewer than 100 citations each. 2009–2011: 66 of 705 flagged (9.4%). 2024–2026: 356 of 15,961 (2.2%). 2026 runs to the data date. Not in the registered plan.
Free to share, with credit.
Download the image, paste the caption, or link to this page. Please credit SaneCite (sanecite.com). The numbers come from the open census data; nothing here is picked by hand.
- How it was read.Three AI models read every citing passage. On a blind sample of 150 citations they agreed with a separate AI reader on whether each cited the paper as support in 142 (94.7%). The rates are not corrected for reader error: corrected using that AI reader's checks, the 40.7% headline estimate is 38.1% (95% CI 33.5–41.8%).
- What a citation is here.One later paper citing one retracted paper, in a year after the retraction. Citations to paperwork retractions, which find no fault in the science, are reported apart.
- What it covers.Open-access papers in Europe PMC citing PubMed-indexed papers retracted from 1990 to 2023. About 43% of all later citations could be read.
- Check it yourself.This chart's data · every citation, searchable · method and limits · code.