More evidence of the corrosive impact of AI on #scholarship #integrity. According to this research in the biomedical context, #fabricated #citations existed in: ~1 in 2,828 papers (2023), ~1 in 458 (2025), ~1 in 277 (early 2026). A twelvefold increase.
Horrifying. The recommendations here are obvious (e.g. #retractions); but I might also suggest that sanctions against authors who are found to have submitted such work are necessary (e.g. desk rejection).
''Black marks on published papers don’t change citation rates, new study finds'' by Retraction Watch
https://retractionwatch.com/2026/04/27/citation-rates-retracted-corrected-articles-asemi-clinical-trials/
#Science #PeerReview #Retractions #Citations #RetractionWatch
Large Language Models (LLMs) can be helpful for literature search and summarisation, but retracted articles can confuse them. This article asks three open weights (offline) LLMs whether 161 high profile retracted articles had been retracted, performing a similar check for a benchmark multidisciplinary set of 34,070 non-retracted articles. Based on titles and abstracts, in over 80% of cases the LLMs claimed that a retracted article had not been retracted (GPT OSS 120B: 82%; Gemma 3 27B: 84%; DeepSeek R1 72B: 88%). The reasons given for a correct retraction declaration were often wrong, even if detailed. This confirms that LLMs have little ability to distinguish between valid and retracted studies, unless they are allowed to, and do, check online. For the benchmark test, there were only 55 false retraction claims from 34,070 non-retracted full text articles, and 28 false claims when only the title and abstract were entered, suggesting that there is only a small chance that LLMs discount valid studies. When retractions are erroneously claimed, this does not seem to be due to mistakes in the article. Overall, the results give new reasons to be cautious about LLM claims about academic findings.
New study: "In over 80% of cases the [tested] #LLMs claimed that a retracted article had not been retracted…LLMs have little ability to distinguish between valid and retracted studies, unless they are allowed to, and do, check online."
https://arxiv.org/abs/2604.16872
Large Language Models (LLMs) can be helpful for literature search and summarisation, but retracted articles can confuse them. This article asks three open weights (offline) LLMs whether 161 high profile retracted articles had been retracted, performing a similar check for a benchmark multidisciplinary set of 34,070 non-retracted articles. Based on titles and abstracts, in over 80% of cases the LLMs claimed that a retracted article had not been retracted (GPT OSS 120B: 82%; Gemma 3 27B: 84%; DeepSeek R1 72B: 88%). The reasons given for a correct retraction declaration were often wrong, even if detailed. This confirms that LLMs have little ability to distinguish between valid and retracted studies, unless they are allowed to, and do, check online. For the benchmark test, there were only 55 false retraction claims from 34,070 non-retracted full text articles, and 28 false claims when only the title and abstract were entered, suggesting that there is only a small chance that LLMs discount valid studies. When retractions are erroneously claimed, this does not seem to be due to mistakes in the article. Overall, the results give new reasons to be cautious about LLM claims about academic findings.
Retraction Watch interview with Elizabeth Bik, who has a preturnatural ability to find duplicated images in different papers:
Will 'AI-Assisted' #Journalists Bring Errors and #Retractions ?
Meet the "journalist" who "uploads press releases or analyst notes into #AI tools and prompts them to spit out #articles that he can edit and #publish quickly," according to the Wall Street Journal. "AI-assisted stories accounted for nearly 20% of Fortune's web traffic in the second half of 2025.
#AiSlop #journalism #journalist
Meet the "journalist" who "uploads press releases or analyst notes into AI tools and prompts them to spit out articles that he can edit and publish quickly," according to the Wall Street Journal. "AI-assisted stories accounted for nearly 20% of Fortune's web traffic in the second half of 2025." An...
Interesting case in which the usual issues surrounding the retraction of a journal article are further complicated by a NY state law criminalizing the distribution of information that endangers health.
https://retractionwatch.com/2026/04/02/judge-lawsuit-controversial-adolescents-paxil-study-329/
PS: Apart from the specific issues raised by this case (whether a certain antidepressant is safe and effective for teenagers), I'm interested in this more general question: Do US journals have a #FirstAmendment right not to retract articles, regardless of the circumstances, for example, even when the articles have been shown to be false and harmful and even when local law prohibits the distribution of such information? Is "forced retraction" a kind of #censorship prohibited by the Constitution?
Before you comment, note that this cases raises a lot of important questions about our tolerance for harm. But I'm trying to raise a slightly different question about the interaction of scholarly norms and public law.
📄🔍 Zurückgezogene Artikel zitieren, ohne es zu merken?
Retractions rechtzeitig zu erkennen ist gar nicht so einfach. Allein 2023 wurden weltweit mehr als 10.000 wissenschaftliche Artikel zurückgezogen. Mit Retraction Watch und dem Zotero-Plugin behält man den Überblick und erkennt Retractions direkt in der eigenen Literaturverwaltung.
#RetractionWatch #Retractions #Zotero #GuteWissenschaftlichePraxis #OpenScience #Openness
https://www.tub.tuhh.de/blog/2026/02/23/retraction-watch-zurueckgezogene-artikel/
George Macgregor
Katharine O'Moore-Klopf, ELS
HO• ⏚
Peter Suber
petersuber
RDN
PrivacyDigest
Dorothy Bishop
Universitätsbibliothek TUHH