Who keeps citing retracted work?
The retracted papers still cited most often as support, and the journals and authors whose papers do it most. Each entry was checked again before publication, as the box below sets out. Census data as of 14 Oct 2026; corrections through 28 Oct 2026.
How these lists were checked
- Papers
- Every citation counted as support was read again in full (23 Oct 2026) for each listed paper and each paper near the last place, 21 papers in all. For the 8 nearest the last place, each citation was read twice, independently, and a third reading settled any disagreement. The 2 papers left off after reading were judged on their second reading, not read in full.
- Journals
- A random sample of 7 to 10 support citations per journal was read again, not every citation.
- Authors
- Every support citation in every listed author's papers was read again (217 in all).
All of these are AI readings.
Everyone named was written to before publication, sent the evidence and asked to tell us if a reading is wrong. Replies received by 28 October are summarised and quoted beside their entries; statements written for publication are printed as written (5 on the page so far; all replies).
The full ranked tables
Every retracted paper (11,697) and every citing journal (3,822) in the lists' scope, ranked by the citations the census AI models labelled as support, with this page's corrections applied. Only the listed entries and the papers near the last place were read again. One held paper is left out until it is named.
All 11,697 papers (CSV) · All 3,822 journals (CSV) · What the tables hold
Retracted papers
The retracted papers most often still cited as support.
Ranked by how many citations, made in a year after the retraction, use the paper as support for a claim.
| # | Retracted paper | Cited as support after retraction | Second reading |
|---|---|---|---|
| 1 | Visfatin: a protein secreted by visceral fat that mimics the effects of insulinFukuhara A et al., Science 2004. Retracted 2007: An Osaka University research integrity committee examined the insulin-receptor experiments and the university recommended retraction; the authors "continue to stand by our conclusions" but found "not all preparations of visfatin bind to and activate the insulin receptor".
| 104of 314 judged (33.1%) | 9 of 14its findings |
| 2 | Cardiovascular Disease, Drug Therapy, and Mortality in Covid-19Mehra MR et al., N Engl J Med 2020. Retracted 2020: The authors could not get the raw data audited, so they were "unable to validate the primary data sources".
| 80of 365 judged (21.9%) | 9 of 12its findings |
| 3 | Predictive validity of a medication adherence measure in an outpatient settingMorisky DE et al., J Clin Hypertens (Greenwich) 2008. Retracted 2023: An independent statistical review found the results "misleading due to issues regarding the sensitivity and specificity" of the adherence measure.Most citations use the MMAS-8 adherence questionnaire and its cut-offs.
| 56of 183 judged (30.6%) | 11 of 14its questionnaire |
| 4 | Association Between Physician Burnout and Patient Safety, Professionalism, and Patient Satisfaction: A Systematic Review and Meta-analysisPanagioti M et al., JAMA Intern Med 2018. Retracted 2020: A University of Manchester panel "found no evidence of intentional fabrication" but, because of flaws in the review process, "cannot confirm that the results of the meta-analysis are wholly valid".
| 53of 203 judged (26.1%) | 11 of 13its findings |
| 5 | The muscle-specific microRNAs miR-1 and miR-133 produce opposing effects on apoptosis by targeting HSP60, HSP70 and caspase-9 in cardiomyocytesXu C et al., J Cell Sci 2007. Retracted 2011: The authors retracted it over "the misuse and re-use of western blot bands" in Figures 3 and 4A, which broke the journal's editorial policy.
| 51of 98 judged (52.0%) | 11 of 14its findings |
| 6 | Increased muscle PGC-1alpha expression protects from sarcopenia and metabolic disease during agingWenz T et al., Proc Natl Acad Sci U S A 2009. Retracted 2016: A University of Cologne investigation "concluded that there was scientific misconduct": blots in Figs. 2D and 3B also appear in an unrelated paper, and a tubulin blot is duplicated.
| 31of 60 judged (51.7%) | 13 of 13its findings |
| 7 | circRNA_100290 plays a role in oral cancer by functioning as a sponge of the miR-29 familyChen L et al., Oncogene 2017. Retracted 2019: The authors found "serious errors in Figs. 5c and 7b, resulting from the data collection process" and are no longer confident in the conclusions.
| 30of 94 judged (31.9%) | 9 of 13its findings |
| 8 | Spontaneous human adult stem cell transformationRubio D et al., Cancer Res 2005. Retracted 2010: The authors could not reproduce some of the reported transformation events and "suspect the phenomenon is due to a cross-contamination artifact"; five of the seven authors agreed to the retraction.
Reply from the senior author, received 22 Oct 2026Dr Bernad says the 2005 paper was the first report of an apparent spontaneous transformation of mesenchymal stem cells in culture, at a time when these cells were being tested in many cell therapies, so it had a large impact on the field. Soon after publication his group found that the main conclusion could not be reproduced and was most probably due to a very small number of contaminating transformed cells that overgrew the cultures. He says the group contacted Cancer Research to offer a retraction; the journal preferred a short notice to the longer one they wrote (de la Fuente et al., 2010). He says the group approached the journal soon after finding the result could not be reproduced. He did not give the date of that contact by 28 October (the notice was published in August 2010). He proposed a note for later authors, printed below as he wrote it, and points to his group's later work on long-term culture of these cells. Soon after publication (Rubio et al., 2005) we realized that the main conclusion was non reproducible and that the putative spontaneous transformation was most probably due to a highly minoritarian number of a transformed cell line.Statement from the senior author, printed as written: Our response: The retraction notice was published in Cancer Research in August 2010 (doi:10.1158/0008-5472.CAN-10-2451). His group's later papers: Estrada JC et al., Cell Death Dis 2013 (replicative senescence of these cells in long-term culture is closely linked to aneuploidy) and Sepúlveda JC et al., Stem Cells 2014 (senescence abrogates their therapeutic potential in a mouse model of lethal endotoxemia). The entry stands: the citations counted were made after the retraction and use the paper as support.Changes since our letter:
| 26of 131 judged (19.8%)each one read twice | 12 of 14its findings |
| 8 | Held: a paper rose to this place after the 23 October corrections. It will be named here once its authors have been written to and have had the same time to reply as everyone else. | ||
| 10 | Selective killing of cancer cells by a small molecule targeting the stress response to ROSRaj L et al., Nature 2011. Retracted 2018: Problems with Fig. 1d and Supplementary Fig. 31b, and missing original data, "raise concerns regarding the integrity of the figures"; five authors disagree with the retraction.
| 25of 93 judged (26.9%)each one read twice | 13 of 15its findings |
Second reading: of up to 15 citations drawn at random and read again (fewer where a later full-text check took one out of scope), how many cite the paper as support, and what they use it for. The sample was drawn before the full read; where the full read changed a citation's label, the final label is the one counted here and in second-reading.csv. Ties share a place. Each reason quotes or closely follows the retraction notice; where we have not yet read a notice, we say so.
Every paper on this list, and each paper near the last place, was read in full on 23 October: each citation counted as "cited as support" was checked against the citing paper's full text, and those that turned out to be passing mentions, warnings about the retraction, citations of the retraction notice or of something else, or that credit the paper with something it does not report, no longer count. For the papers at and just below the last places, each citation was read twice, independently, and a third reading settled the 11 on which the first two disagreed. These are AI readings. The papers left off after reading were judged on their second reading, not read in full.
Left off after reading
These papers would make the list on the count alone, but reading them showed the citations cite something the retraction does not touch, for example a tool or method in the paper that still works.
| Retracted paper | Support | Why it is left off |
|---|---|---|
| An enhanced transient expression system in plants based on suppression of gene silencing by the p19 protein of tomato bushy stunt virusVoinnet O et al., Plant J 2003. Retracted 2015: Image duplications in Figure 3b and in loading-control lanes of Figures 2g, 3a, 4e and 4f; the authors "accept that integrity of the scientific literature is compromised by the data manipulation". | 167of 246 | Citations use the p19 lab tool the paper describes, which still works (14 of 15 read). It was retracted over duplicated figures. |
| Global Increase in Breast Cancer Incidence: Risk Factors and Preventive MeasuresKashyap D et al., Biomed Res Int 2022. Retracted 2023: The publisher found "evidence of systematic manipulation of the publication and peer-review process" and cannot "vouch for the reliability or integrity" of the article. | 29of 143 | A review: citations use its summary figures, which come from other published sources. |
Citing journals
The journals whose papers most often cite retracted work as support.
Two ways to rank: the count favours large journals, the share is fairer to small ones. The census reads open-access papers, so journals that publish everything open access are counted in full and others in part: large open-access journals rank high on counts partly because more of their papers could be read. Compare journals by the share table. These are the journals that published the citing papers, not the retracted ones.
Most citations of retracted work as support
| # | Citing journal | Cited as support | Second reading |
|---|---|---|---|
| 1 | Int J Mol Sci610 papers cite retracted work as support | 654654 of 1,547 judged (42.3%) | 9 of 9 |
| 2 | Sci Rep570 papers cite retracted work as support | 582582 of 1,731 judged (33.6%) | 7 of 10 |
| 3 | PLoS One487 papers cite retracted work as support | 497497 of 1,586 judged (31.3%) | 7 of 8 |
| 4 | Front Pharmacol320 papers cite retracted work as support | 348348 of 680 judged (51.2%) | 8 of 10 |
| 5 | Front Immunol299 papers cite retracted work as support | 326326 of 720 judged (45.3%) | 10 of 10 |
| 6 | Front Oncol269 papers cite retracted work as support | 313313 of 629 judged (49.8%) | 8 of 91 of the papers read again has since been retracted (PMC10580988) |
| 7 | Heliyon276 papers cite retracted work as support | 298298 of 650 judged (45.8%) | 7 of 91 of the papers read again has since been retracted (PMC11033103) |
| 8 | Cancers (Basel)209 papers cite retracted work as support | 238238 of 544 judged (43.8%) | 6 of 9 |
| 9 | Molecules191 papers cite retracted work as support | 203203 of 488 judged (41.6%) | 8 of 10 |
| 10 | Front Cell Dev Biol134 papers cite retracted work as support | 163163 of 295 judged (55.3%) | 8 of 8 |
Highest share as support, among journals with 100+ judged citations
| # | Citing journal | Share cited as support | Second reading |
|---|---|---|---|
| 1 | Front Mol Biosci57 papers cite retracted work as support | 60.3%70 of 116 judged (60.3%) | 9 of 9 |
| 2 | Noncoding RNA Res42 papers cite retracted work as support | 58.3%67 of 115 judged (58.3%) | 9 of 9 |
| 3 | Cell Commun Signal54 papers cite retracted work as support | 58.2%64 of 110 judged (58.2%) | 8 of 8 |
| 4 | Front Cell Dev Biol134 papers cite retracted work as support | 55.3%163 of 295 judged (55.3%) | 8 of 8 |
| 5 | Cancer Cell Int116 papers cite retracted work as support | 55.2%143 of 259 judged (55.2%) | 7 of 9 |
| 6 | Onco Targets Ther62 papers cite retracted work as support | 53.9%62 of 115 judged (53.9%) | 6 of 71 of the papers read again has since been retracted (PMC6885654) |
| 7 | Bioengineered77 papers cite retracted work as support | 52.9%83 of 157 judged (52.9%) | 8 of 8 |
| 8 | Front Pharmacol320 papers cite retracted work as support | 51.2%348 of 680 judged (51.2%) | 8 of 10 |
| 9 | Discov Oncol92 papers cite retracted work as support | 51.0%106 of 208 judged (51.0%) | 8 of 9 |
| 10 | Front Neurosci48 papers cite retracted work as support | 50.5%51 of 101 judged (50.5%) | 8 of 9 |
Second reading: 10 support citations per journal, drawn at random and read again; citations of work retracted for a reason that does not touch the results were later taken out of the samples. Across the lists' scope, 40.3% of citations are support (the census headline, which also counts papers PubMed does not yet link to a notice, is 40.7%). With 100 to 300 citations, a share can be off by several points either way, so close places in the second table are not real differences.
Citing authors
The authors whose papers most often cite retracted work as support.
This counts citations, not intent. It does not say anyone acted dishonestly, and it does not judge whether a review's conclusions still stand: most of these citations sit in review articles that sum up many earlier studies, often in a table, and a co-author may have written that part.
| # | Author | Papers citing retracted work as support | Rate | Reviews | Retracted papers cited |
|---|---|---|---|---|---|
| 1 | Mohammad TaheriORCID 0000-0001-8381-0591Co-wrote 22 of them with Bashdar Mahmud Hussen and 21 of them with Soudeh Ghafouri-Fard | 24of 26 found by the census | 10.5%of 228 tagged papers | 23 | 34 |
The 24 papers and 34 citations
The published reference list marks the work as retracted
| |||||
| 2 | Bashdar Mahmud HussenORCID 0000-0003-1060-6527Co-wrote 22 of them with Mohammad Taheri and 19 of them with Soudeh Ghafouri-Fard | 22of 24 found by the census | 17.7%of 124 tagged papers | 21 | 32 |
The 22 papers and 32 citations
The published reference list marks the work as retracted
| |||||
| 2 | Soudeh Ghafouri-FardORCID 0000-0002-0223-499XCo-wrote 21 of them with Mohammad Taheri and 19 of them with Bashdar Mahmud Hussen | 22of 24 found by the census | 8.2%of 268 tagged papers | 21 | 28 |
The 22 papers and 28 citations
The published reference list marks the work as retracted
| |||||
| 4 | Michael R HamblinORCID 0000-0001-6431-4605 | 9of 9 found by the census | 4.9%of 184 tagged papers | 9 | 34 |
The 9 papers and 34 citations
The published reference list marks the work as retracted
Reply from the author, received 22 Oct 2026Dr Hamblin says he has never knowingly cited a retracted paper and finds it hard to believe that anyone would. The papers in our letter all came from his collaborators in Iran, where internet access is sporadic, unreliable and slow, so researchers build a reference library when they have access and write from it later; a paper retracted after the library was built would not show as retracted, which he says explains our findings. He said none of these citations contains the word "retracted", though our letter said one reference list marked a work retracted. I have never knowingly cited a retracted paper and frankly I find it difficult to believe that anybody would knowingly do thisOur response: He was right about the one item our letter marked: its reference carries a link to the retraction notice, most likely added during publication. For every author, such items now sit apart under their own heading, with a note that publishers often add these links and labels, so they do not show that the authors knew. The other citations stay counted, because the reviews were published after the retractions.Changes since our letter:
| |||||
| 5 | Javad Sharifi-RadORCID 0000-0002-7301-8151 | 8of 8 found by the census | 3.1%of 262 tagged papers | 8 | 8 |
The 8 papers and 8 citations
The published reference list marks the work as retracted
| |||||
| 6 | Abdelhakim BouyahyaORCID 0000-0001-9317-1631 | 7of 8 found by the census | 4.9%of 142 tagged papers | 7 | 9 |
The 7 papers and 10 citations
The published reference list marks the work as retracted
| |||||
| 7 | Anupam BishayeeORCID 0000-0001-9159-960X | 6of 8 found by the census | 6.9%of 87 tagged papers | 4 | 6 |
The 6 papers and 6 citations
The published reference list marks the work as retracted
| |||||
| 7 | Meysam MoghbeliORCID 0000-0001-9680-0309 | 6of 7 found by the census | 18.2%of 33 tagged papers | 6 | 10 |
The 6 papers and 10 citations
Reply from the author, received 22 Oct 2026In two replies, Dr Moghbeli confirms that the citations are correct. He says each review was finished and first submitted before the cited papers were retracted. In his second reply he adds that the six reviews were submitted, rejected and passed from journal to journal over long periods, often through Springer's transfer to the next journal it suggested, and that the retractions came during those transfers. Once a manuscript is under review, authors cannot change it unless the journal sends it back, and journals did not then check references for retractions as they do now; so he says authors cannot find such references unless the journal checks them and asks for a correction, and that such citations are inevitable in reviews with hundreds of references that pass through several journals. He says the retracted references, 11 of the 1,546 in the seven reviews, have no significant negative effect on the reviews' topics or conclusions. He also said the first item, Roberts JT et al. (Ann Oncol 2006), is not retracted and has no retraction notice. When we wrote, prepared, and submitted these review manuscripts, the cited papers were not yet retracted.Our response: The record he checked is the original 2005 trial in J Clin Oncol, which is not retracted; the Ann Oncol 2006 paper was retracted in 2011 as a redundant publication of it, and had already been left out under the rule for duplicates. His timelines match the journal records and the notices and are printed beside each review; those citations stay counted, because the reviews were published after the retractions. His second reply explains how the manuscripts moved between journals; it changes no entry.Changes since our letter:
| |||||
| 7 | Amirhossein SahebkarORCID 0000-0002-8656-1444 | 6of 6 found by the census | 4.2%of 144 tagged papers | 6 | 9 |
The 6 papers and 9 citations
The published reference list marks the work as retracted
Reply from the author, received 21 Oct 2026Dr Sahebkar says a retraction is not a single verdict on a paper: notices range from fabricated data to duplicate publication, authorship disputes or a problem in one figure unrelated to the result being cited, and COPE guidance recommends citing retracted work with its status clearly marked. Whether a citation is a problem depends on whether the reason for retraction touches the statement it supports and whether the citing paper's conclusion rests on it, which a count of citations cannot answer. All seven papers in our letter are reviews of hundreds of sources, and he says none of their conclusions depends on the retracted papers. He asks that authors be compared by rates normalised by their total publications and cited references, not ranked by absolute counts, since a ranking by count is led by the authors who publish most. He was given the opportunity to send a direct response for publication instead of this summary.Our response: Two of his points changed the lists for every author: the author list now shows a rate beside each count (for him, 6 of 144 papers, 4.2%), and citations of work retracted for reasons that do not touch its results are left out. The ranking stays by count, as set out in advance, and the rate counts papers, not references.Changes since our letter:
| |||||
| 7 | Noushin NabaviORCID 0000-0002-5471-6229 | 6of 7 found by the census | 16.7%of 36 tagged papers | 6 | 10 |
The 6 papers and 10 citations
Reply from the author, received 21 Oct 2026Dr Nabavi checked the references and passages and accepts that the cited studies have retraction notices and are cited in support of statements in her reviews, but says this does not by itself show prior awareness, misconduct, or that the reviews' overall conclusions are invalid. She notes that her oxaliplatin review was submitted on 21 September 2023, before the cited circPTK2 study was retracted on 18 October 2023, though it was revised and published afterwards. She would need to review the manuscript history before giving an account of awareness, and is willing to work with her co-authors and the journals to assess the affected statements and pursue appropriate corrections. She asked that her entry be deferred until she has seen and answered the human check's findings, and that her response be published alongside her entry; it is printed below as she wrote it. Citation counts alone cannot establish that an entire review is unreliable.Statement from the author, printed as written: Our response: Her dates are exact, and the timeline is printed beside that citation; it stays counted, because every entry is measured by the paper as published. We did not defer her entry, since the list is published whole. If she shows that a reading is wrong, we correct the entry and say so here.Changes since our letter:
| |||||
| 7 | Gautam SethiORCID 0000-0002-8677-8475 | 6of 7 found by the census | 3.1%of 195 tagged papers | 5 | 7 |
The 6 papers and 7 citations
| |||||
| 7 | Chih-Yang HuangORCID 0000-0003-2347-0411 | 6of 6 found by the census | 2.9%of 204 tagged papers | 1 | 6 |
The 6 papers and 6 citations
| |||||
| 7 | Peng LuoORCID 0000-0002-8215-2045 | 6of 6 found by the census | 2.3%of 259 tagged papers | 3 | 6 |
The 6 papers and 6 citations
The published reference list marks the work as retracted
| |||||
Papers: citing papers with at least one citation the second reading confirmed as support. Rate: those papers as a share of the author's open-access papers in PMC that carry their ORCID iD in Europe PMC, 2010 to 2026 (counted 2026-10-21). The ranking stays by count. Below the top three, the order does not hold under rates: authors with few tagged papers can have high rates on small numbers. Reviews: how many of those papers Europe PMC lists as review articles. Every support citation in every listed author's papers was read again (217 in all). The 3 authors who share papers wrote most of them together, so their counts overlap. Ties share a place.
Each listed item names the notice's reason. Results: the notice names a problem with the data, images or results, or says they cannot be relied on. Unclear: it gives no reason or a vague one, or we have not read it yet; these still count. Citations of work retracted only for reasons that do not touch the results are left out, as the census leaves out paperwork retractions. Where a citing paper was submitted or accepted before the retraction, its timeline is printed beside the item; papers that went online before the retraction are left out. 22 of the 200 items sit apart: the citing paper's published reference list marks the work as retracted. Such labels and links are often added automatically during publication, so they do not show that the authors knew.
How we checked
Counted by the census, then read again.
- AI readings.Each label is our reading of a published passage by the method on this page, and each paper and author item links to its passages so you can judge it yourself. The counts come from the Phase 2 census: three AI models read every citing passage. Each list was then checked again, as set out in the box at the top; those are AI readings too. Everyone named was sent the evidence and asked to tell us if a reading is wrong; where one is, we correct the entry and say so beside it.
- Dates and reasons from the notice.Each retraction date and reason on the lists comes from the journal's retraction notice, not from a database. Where we have not yet read a notice, we say so instead of giving a reason. Corrections made after the data was built are listed, with their sources, in list_corrections.json.
- What each paper reached.For every paper the census found citing a listed paper as support after its retraction, we read its Europe PMC record on 29 October 2026: its publication type (review, systematic review, trial), how often it has been cited since, and the grants it acknowledges. "Cited in turn" counts every later citation of those papers, not only citations for this claim, so it measures reach, not harm. Totals are in reach.json.
- Only citations made after the retraction.A citation counts if the citing paper came out in a year after the retraction. Where we checked dates one by one, papers that went online before the notice are left out, and papers submitted or accepted before it carry their timeline. The rest of the census goes by calendar year, so the full tables may still hold a few citations made before a retraction.
- Only retractions about the results.Citations of a paper are left out when its retraction notice gives only reasons that do not touch the results, such as reused text, a duplicate of a result that stays published, authorship, ethics paperwork, copyright or a software licence, or a wrong journal; the census already leaves out paperwork retractions the same way. We apply this where we have read the notice. A notice that gives no reason, or an unclear one, still counts and is marked "reason unclear". They are not hidden: 289 later citations use 20 such papers as support, from 289 citing papers. The papers · every citation, labelled with the reason.
- Only confirmed retractions.The lists use only papers PubMed links to a retraction notice: 45,763 judged citations to 11,698 retracted papers. Papers replaced by a corrected version, and paperwork retractions that find no fault in the science, are left out.
- Authors by ORCID only.Authors are matched by their ORCID iD, so two people with the same name are never merged. Authors without an ORCID iD in Europe PMC cannot be listed, so this is not every author.
- What the second reading found.Of 601 different citations the census called support and we read again, 512 (85.2%) held up after every reading. 59 only mentioned the paper in passing, 7 flagged it as retracted or unreliable, and 23 were about something else, most often because the reference pointed at the wrong paper. These citations come from the top of each list, so they are not a rate for the whole census. Some papers were cited only for a tool or method that still works; those are left off, with the reason shown.
- Open-access papers only.The census reads papers whose full text is open in Europe PMC. Journals that publish everything open access are counted in full, others only in part, so the counts favour large open-access journals.
- Check it yourself.Each list as a CSV: papers · journals by count · journals by share · authors. every author item, with the notice's reason class and any timeline. Every citation read again, with its verdict and a note. Every reply from the people named, as summarised. What each column means.
- The full tables.All 11,697 retracted papers and all 3,822 citing journals. They are ranked by count with the corrections in list_corrections.json applied: 57 papers' counts differ from the census labels alone, and the file gives both. Only the listed entries and the papers near the last place were read again. One held paper is left out until it is named. All papers · all journals · PubMed retraction status of every paper. Their reason column is grouped from Retraction Watch codes, not the notice's wording.
- Corrections.If you are named and think an entry is wrong, write to SaneCite at letters@sanecite.com, with the citing paper's PMC number if you have it. We reply within two working days and decide within seven days. A correction is made beside the entry, dated, and logged in list_corrections.json. You can also ask us to take an entry down: it comes off when the evidence shows a reading is wrong or a citation is out of scope, the same rule for everyone, and if it stays we tell you why.
Named people and your data
What we hold about the people named, and why.
- Who is responsible.SaneCite, an independent research project in Georgia, USA. To reach us, write to SaneCite at letters@sanecite.com.
- What we use.Names, ORCID iDs and publication records: titles, journals, dates, reference lists and the passages that cite retracted work. And the replies people send us.
- Where it comes from.Public records: Europe PMC (with the ORCID iDs it holds), PubMed, the Retraction Watch database and the journals' own retraction notices.
- Why, and on what basis.Research in the public interest on how the published record handles retracted work, and publishing its results. Where UK or European data protection law applies, we rely on legitimate interests in research and free expression, and on the exemptions those laws give to academic and journalistic work. Entries come from whole-set counts, never picked by hand.
- How entries are made.By AI readings of the citing passages, by the method on this page, with a right of reply for the people named before publication.
- What we print from a reply.A summary in our words, one short quote, and any statement written for publication, the same for everyone. We never print email addresses, phone numbers or details about other people. The full reply is kept privately and shared only when needed, with a journal, a fiscal sponsor or a lawyer.
- How long.Published entries stay as part of the research record, with every correction dated. Private correspondence is kept while the entry is published and for two years after.
- Your rights.You can ask us to correct your entry, to tell you what we hold about you, to restrict or stop a use, or to delete what we hold: write to SaneCite at letters@sanecite.com. We reply within two working days and decide within one month. Where we keep a published entry because the research record needs it, we tell you why. You can also complain to your data protection authority (in the UK, the Information Commissioner's Office).