Research integrity and peer review最新文献

{"title":"Possibilities for the peer review of interdisciplinary research: a small-scale qualitative study.","authors":"Georgia Vesma","doi":"10.1186/s41073-026-00204-3","DOIUrl":"10.1186/s41073-026-00204-3","url":null,"abstract":"<p><strong>Background: </strong>Despite critiques from across academia, double-anonymised peer review is still considered the 'gold standard' quality-assurance tool for the publication of academic research. While some journals make use of 'open' peer-review practices to make elements of academic publishing more transparent, anonymised is still used by the majority of academic journals, with anonymity considered the best guarantor of reviewer objectivity. Anonymised peer review remains a 'black box' which conceals diverse and idiosyncratic practices by reviewers, editors and journals. This paper seeks to uncover the impact of such practices on the peer review of interdisciplinary research.</p><p><strong>Methods: </strong>The author first conducted a survey of 164 self-identified interdisciplinary researchers from a range of disciplines at one research-intensive UK university to identify barriers to and enablers of high-quality peer review for interdisciplinary papers. This quantitative/qualitative survey was supplemented by interviews with authors, editors and reviewers from across biological and medical sciences, physical sciences and engineering and social sciences and humanities.</p><p><strong>Results: </strong>Authors of interdisciplinary research papers reported experiencing disadvantage in peer review practices, including perceived bias against interdisciplinary research by mono-disciplinary reviewers and the assignment of unsuitable reviewers. Reviewers reported being assigned papers to review that were not within their field of expertise, and insufficient editorial guidance on review processes for interdisciplinary papers. Editors reported that the 'reviewer crisis' has further complicated the already challenging process of seeking peer review for interdisciplinary research. In interviews, reviewers shared strategies to overcome weaknesses in the review process for interdisciplinary papers. These included confidence reporting where the reviewer did not feel able to comment on all aspects of the paper, and seeking advice from colleagues to enhance their review. Practices were employed with varying degrees of formality, and in some cases undermined the stated goals of anonymised peer review.</p><p><strong>Conclusions: </strong>Considering the possible implications of the informal strategies employed by peer reviewers, the paper proposes a framework for interdisciplinary peer review informed by existing open peer review practices that formalises and incorporates existing strategies used by reviewers and editors to enhance the peer review of interdisciplinary research papers.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13270611/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148260757","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Statistical reviewing for a medical journal: summary of workshop proceedings.","authors":"Derrick Bennett, Sam Leary, Paula Dhiman, Jen Lewis, Julie Morris, Gordon Prescott, Andy Vail, Ashma Krishan, Antonia Marsden, Clare Robinson, Jamie Sergeant","doi":"10.1186/s41073-026-00227-w","DOIUrl":"10.1186/s41073-026-00227-w","url":null,"abstract":"<p><p>The use of statistical methods in medical research has long been a concern and careful review of these methods is required to ensure that published papers are of sufficient statistical quality. However, there is a shortage of statistical reviewers for medical journals and there is a dearth of training for those with little or no experience. An online workshop on statistical reviewing for medical journals was convened by the Improving Statistical Literacy and Early Career Development streams of the National Institute for Health and Care Research (NIHR) Statistics Group to address this gap. This article summarizes the key learnings from this workshop and provides some practical advice for those potentially interested in undertaking statistical reviews for a medical journal.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13267749/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148254879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Irreproducible research and a typology of replication efforts.","authors":"Adrian Treves, Sedona Chinn, Miha Krofel","doi":"10.1186/s41073-026-00203-4","DOIUrl":"10.1186/s41073-026-00203-4","url":null,"abstract":"<p><p>The scholarly and scientific literature does not automatically correct itself. Erroneous findings may persist without correction or retraction. Following prior definitions of zombie articles that are retracted but continue to be cited affirmatively, we add another category of 'undead' articles. We define 'vampire' articles by two jointly necessary conditions. The first condition is irreproducibility, demonstrated by one or more failed efforts at replication or conditions that make replication impossible. We offer a novel typology of four categories of replication efforts. We propose a rule of thumb for how many failed efforts at replication might be required for each type of replication effort before the original finding is deemed irreproducible. The second necessary condition for a putative vampire article is that it is cited affirmatively in public policy, the scholarly literature, or private (neither governmental, nor published in the peer-reviewed scholarly literature) communications, after the first condition is met. We discuss rules of thumb for how many such affirmative citations might lead qualified researchers in that subfield to propose correction, retraction, or editorial note of concern for the article in question. Our first case concerns aerial shooting at coyotes and the second case predicted over-fishing. We discuss the damaging effects of vampire articles and why the metaphor has heuristic value and utility. Perhaps the current retraction guidelines could be sharpened to include considerations around vampire articles. We also discuss lessons from the communication sciences about how to remedy misinformation, offering recommendations to researchers, publishers and editors who are concerned with correcting their literatures and public trust in science generally.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13267231/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148254901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Data integrity of the 223 randomized controlled clinical trials produced by multiple groups centered around a single author.","authors":"Daan Langeveld, Buğra Sipahioğlu, Marlena Oebel, Steven Miletić, Roy de Kleijn, Marc Molendijk","doi":"10.1186/s41073-026-00224-z","DOIUrl":"10.1186/s41073-026-00224-z","url":null,"abstract":"<p><strong>Background: </strong>We suspected unusual consistency in baseline variables and dropout rate in the randomized controlled trials (RCTs) produced by a team of researchers. We define this group of RCTs by the presence of an individual researcher in the author team, whom we call here the 'key author'. We investigated the trustworthiness of these data and with that its scientific and clinical value.</p><p><strong>Methods: </strong>Systematic searches were performed in Embase, PubMed, and Web of Science for all the RCTs published by this author team. The final search date was 01/08/2024. Preregistered analyses tested for the presence of sampling variability in baseline characteristics of the RCTs and participant flow, operationalized as an even vs. odd dropout and an equal dropout, through the trials. Control data was gathered from RCTs that were performed by other groups but on similar topics as the investigated RCTs.</p><p><strong>Results: </strong>We found strong evidence that the expected sampling variability was not present in the 223 RCTs produced by this group relative to the expected uniform distribution (Bayes Factor [BF₁₀] > 100000) and to what we observed in the 103 control RCTs (BF₁₀ = 172). The proportion of even vs. odd dropout (e.g., 4 vs. 3) was .76 (95% CI = 0.70, 0.81) and different from the theoretical value of .50 (BF₁₀ = 8.5×10¹¹) and from what we observed in control RCTs (BF₁₀ = 379). Equal dropout over arms of a trial was 3 to 4 times more likely in the tested trials, relative to what we observed in control trials (BF₁₀ = 1.8×10¹⁶). The results were not explained by the method of randomization or sample size. Associations were mostly evident when the 'key author' had first or last authorship.</p><p><strong>Conclusion: </strong>It is unlikely that the RCTs by this team were generated through the process of randomization, and the flow of participants in these trials seems implausible. Hence, we question the scientific and clinical value of the RCTs that are produced by this team.</p><p><strong>Registration: </strong>https://doi.org/10.17605/OSF.IO/8MTG6.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":" ","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13584280/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148254862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Making scientific publishing more balanced and reversing the current dangerous drift to better support global environmental conservation and planetary health.","authors":"Paolo Guidetti, Carlo Nike Bianchi, John Baxter, Nane Pelke, Charles-François Boudouresque","doi":"10.1186/s41073-026-00206-1","DOIUrl":"10.1186/s41073-026-00206-1","url":null,"abstract":"<p><p>The paper examines structural distortions in contemporary scientific publishing and their implications for research quality, equity, ethics and environmental conservation. Despite being largely publicly funded, scientists continue to provide unpaid labour as reviewers and editors while facing increasing publication fees or subscription costs. The rapid proliferation of journals has intensified reviewer fatigue, reduced the availability of qualified referees, and contributed to declining peer-review quality. These conditions, combined with status biases and insufficient editorial oversight, seriously risk to undermine the reliability of published research. The expansion of predatory journals and exploitative open-access models further erodes trust in scientific communication. In a publish-or-perish environment driven by bibliometric evaluation, there is a proliferation of low-quality and fraudulent research, amplified by AI-generated content. This dynamic benefits commercial publishers, whose profit margins have grown sharply, while draining financial resources from academic institutions. The resulting \"vampirization\" of the research system exacerbates global inequities and contributes to an exponential increase in publications with limited scientific or societal impact. To counteract this trajectory, we propose that scientists prioritise journals governed by scientific societies or public institutions, and those that adopt ethical publishing practices, especially avoiding predatory publishers. Collective actions-such as mass resignations from editorial boards in response to unreasonable publication charges-are presented as effective strategies. Institutions are encouraged to shift evaluation criteria from quantity to quality, discourage unethical practices, and promote interdisciplinary collaboration, particularly in conservation science. We acknowledge the inherent imperfections of past systems but emphasise that the current drift poses significant risks, especially for fields informing environmental policy making. Declining primary research quality may cascade into misguided decision-making at all scales. We call for systemic reform, including an international accreditation system for journals and publishers based on transparent ethical standards and overseen by public research agencies. Ensuring access to reliable, high-quality scientific information is essential for supporting conservation, sustainability, and the broader societal role of science.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13262215/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148229567","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How do PhD candidates perceive good research practices in the Netherlands Code of Conduct for Research Integrity?","authors":"Lodewijk A Pet, Joeri K Tijdink, Anna E van 't Veer, Ionica Smeets, M Yvonne H G Erkens, Frits R Rosendaal, Bob Siegerink","doi":"10.1186/s41073-026-00201-6","DOIUrl":"10.1186/s41073-026-00201-6","url":null,"abstract":"<p><strong>Introduction: </strong>Research misconduct and questionable research practices compromise the quality of scientific research. The Netherlands Code of Conduct for Research Integrity 2018 outlines five core principles and 61 standards for good research practices intended to guide researchers and research integrity committees. For the code to function, its 61 standards must be perceived as clear and relevant by those who perform the research practices. These include researchers including PhD candidates, who contribute about 40% of Dutch research full-time equivalents. This study examines how PhD candidates in the Science and Medicine faculties at Leiden University perceive these 61 standards.</p><p><strong>Methods: </strong>A total of 332 PhD candidates (73% participation proportion) evaluated the standards in the Dutch code as part of a mandatory research integrity course at Leiden University and consented to share their responses for research purposes. Each participant rated half of the standards using Likert-type scales for clarity, relevance, frequency, and seriousness of non-adherence. In addition, open-text responses were collected and categorized according to the code's five core principles.</p><p><strong>Results: </strong>Overall, 83% of clarity ratings and 77% of relevance ratings scored 4 or 5, indicating a generally positive perception of the standards. However, the perceived frequency of non-adherence varied across standards (5-34%). There were discrepancies in severity assessments, where standards that should represent minor shortcomings were perceived as serious misconduct and the other way around. Open-text responses primarily emphasized honesty and transparency, while independence was rarely mentioned, and participants also highlighted values not explicitly covered by the current code, such as equality and ethical collaboration.</p><p><strong>Discussion: </strong>PhD candidates generally regarded the Dutch Code's 61 standards as clear and relevant, but their ratings of how serious non-adherence should be considered often diverged from the code's a priori three-tier severity framework. Several standards labelled as \"minor shortcomings\", particularly those related to ethics approvals, authorship, and transparent reporting, were frequently classified as research misconduct, while one plagiarism-related standard (practice 34) was more often seen as a questionable practice. These results suggest the need to refine currently ambiguous formulations and reconsider the a priori severity categories.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13255387/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148220776","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Reviewer recommendations and editorial outcomes in peer review: a longitudinal analysis of agreement and disagreement across review rounds.","authors":"Carlos De Las Cuevas","doi":"10.1186/s41073-026-00200-7","DOIUrl":"10.1186/s41073-026-00200-7","url":null,"abstract":"<p><strong>Background: </strong>Peer review is often portrayed as a consensus-driven process in which reviewer recommendations largely determine editorial decisions. However, manuscript-level evidence describing how reviewer disagreement unfolds across review rounds-and how it is resolved in routine editorial practice-remains limited.</p><p><strong>Methods: </strong>We conducted a meta-research analysis of anonymized editorial data from Healthcare (MDPI, Basel, Switzerland; ISSN 2227-9032) (2021-2025), including manuscripts that were ultimately published and manuscripts rejected after external peer review. Reviewer recommendations (accept, minor revision, major revision, reject) were examined at the manuscript level across review rounds. Alignment between reviewer recommendations and editorial outcomes was assessed primarily using the majority recommendation in the final review round, applying a conservative tie-breaking rule (reject > major > minor > accept). Reviewer discordance was defined as the presence of two or more distinct recommendation categories within a review round. Longitudinal trajectories from first to final round were analyzed, with sensitivity analyses using alternative definitions of unfavorable recommendations. The analysis was conducted using anonymized editorial records. The author did not access identifiable reviewer or author information, and the dataset was analyzed in aggregated form to minimize the risk of re-identification.</p><p><strong>Results: </strong>The analysis included 12,187 published manuscripts and 3,819 rejected manuscripts, corresponding to 21,497 and 5,189 review-round records, respectively. Reviewer discordance was common in the first round for both published and rejected manuscripts (74.8% and 85.8%). Across rounds, discordance declined markedly among published manuscripts (to 29.3% in the final round) but remained high among rejected manuscripts (71.4%). Overall alignment between the majority recommendation in the final round and the editorial outcome was high (≈88%) but not absolute: 6.3% of published manuscripts had a majority recommendation of rejection in the final round, and 8.0% retained at least one rejection recommendation. Longitudinally, discordance was frequently resolved among published manuscripts (53.1% transitioned from discordant to concordant), whereas it often persisted among rejected manuscripts (61.4% remained discordant).</p><p><strong>Conclusions: </strong>In routine practice, peer review functions less as a voting mechanism and more as an adjudicative process in which editorial judgment integrates heterogeneous reviewer input. The asymmetric resolution of disagreement across outcomes underscores the central role of editorial decision-making when reviewer recommendations diverge.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":"11 1","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13248283/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148207257","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI text detection in dentistry: a comparative analysis across generative models.","authors":"Jacopo Villa, Daniele Garcovich, Luca Lombardo, Giuseppe Siciliani, Milagros Adobes Martin","doi":"10.1186/s41073-026-00228-9","DOIUrl":"10.1186/s41073-026-00228-9","url":null,"abstract":"<p><strong>Background: </strong>Robust screening for AI-generated scientific text is increasingly required by journals, yet detector performance on full-length manuscripts remains unclear. Six widely used detectors were benchmarked on dentistry manuscripts, and performance was compared across state-of-the-art generators.</p><p><strong>Methods: </strong>A total of 120 manuscripts was assembled in four groups (GPT-4.5, GPT-4o, DeepSeek-R2, and human-written; n = 30 each). Aidetector, GPTZero, Copyleaks, Originality.AI, Turnitin, and DetectingAI produced 0-100% scores. Discrimination was assessed with ROC/AUC and DeLong tests with Bonferroni correction. A pre-specified 60% threshold yielded document-level classifications. Inter-detector agreement was quantified with Cohen's κ.</p><p><strong>Results: </strong>Four detectors showed high discrimination; pairwise AUC differences among these were not significant after Bonferroni correction and each outperformed Turnitin and DetectingAI. Using the 60% cut-off, sensitivities and specificities were high, with no false positives observed among the 30 human texts included in this sample. Agreement was almost perfect among the high performers, slight for Turnitin, and none for DetectingAI. Across tools, DeepSeek-R2 texts were the easiest to detect.</p><p><strong>Conclusions: </strong>On full dentistry manuscripts, four detectors showed high discrimination, whereas Turnitin showed moderate performance and DetectingAI was ineffective. A percentage-based 60% decision threshold provided reproducible, manuscript-level calls.</p><p><strong>Clinical relevance: </strong>These results may help dental editors and reviewers compare detectors for preliminary screening of AI-generated text under controlled conditions.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":" ","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13474814/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148206938","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research integrity and transparency in CAM journals: an analysis of publication bias and reporting practices (1995-2023) in four leading journals.","authors":"Stephen Neilson, Kevin Smith","doi":"10.1186/s41073-026-00221-2","DOIUrl":"10.1186/s41073-026-00221-2","url":null,"abstract":"<p><strong>Background: </strong>Complementary and alternative medicine (CAM) remains widely used worldwide, yet longstanding concerns persist regarding the balance and reliability of the evidence presented in CAM journals.</p><p><strong>Objective: </strong>To examine long-term trends in publication practices within leading CAM journals, with particular attention to changes in publication types and the prevalence of positive versus negative study outcomes as indirect indicators of potential publication bias.</p><p><strong>Methods: </strong>We conducted a complete census of articles published in four leading CAM journals at two contemporary time points (2018 and 2023), replicating the design and classification framework of a seminal 2001 analysis covering 1995 and 2000. Articles were categorised by publication type, subject area, and author-reported study outcome (positive, negative, or inconclusive, corresponding to the \"open\" category used in the original 2001 study). Descriptive analyses were used to compare trends over time and with earlier findings.</p><p><strong>Results: </strong>The total number of published articles increased substantially between the earlier and contemporary periods. The proportion of empirical studies, particularly clinical trials, rose over time. However, the prevalence of positive outcomes also increased markedly, with positive findings accounting for over 80% of published articles in the contemporary period, compared with 49% in the earlier study. Negative and inconclusive outcomes remained relatively infrequent.</p><p><strong>Conclusions: </strong>Despite growth in publication volume and a shift toward empirical study designs, CAM journals continue to exhibit a pronounced predominance of positive outcome reporting. These findings suggest that longstanding concerns regarding publication bias in CAM publishing have not diminished over time and appear to have intensified, with selective publication and related reporting and dissemination practices plausibly contributing to the observed patterns. This has important implications for research integrity and evidence-based decision-making in medical practice.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":" ","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13471484/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148177341","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Linguistic shifts in dental scientific writing in the large language model era.","authors":"Hadar Better, Doron Haim, Thabet Asbi, Gavriel Chaushu, Keren Better","doi":"10.1186/s41073-026-00226-x","DOIUrl":"10.1186/s41073-026-00226-x","url":null,"abstract":"<p><strong>Background: </strong>The integrity of scientific literature depends on transparent authorship and authentic communication of research findings. The release of ChatGPT in November 2022 introduced powerful AI writing assistance that may be changing how dental researchers write manuscripts. However, systematic evidence documenting this shift and its implications for research integrity remains limited. Traditional detection tools and self-reported surveys have proven unreliable, requiring alternative approaches to assess AI's influence on published literature.</p><p><strong>Objective: </strong>To detect and quantify linguistic changes in dental scientific abstracts by comparing pre-ChatGPT (2021) and post-ChatGPT (2025) periods, providing empirical evidence of AI adoption patterns and their implications for scientific communication integrity.</p><p><strong>Methods: </strong>We analyzed 2,770 abstracts from six major dental journals using vocabulary-based linguistic analysis. Following computational linguistics methods validated in biomedical literature, we extracted six AI-associated features and computed composite detection scores. This observational text-based approach avoids reporting bias while providing quantifiable evidence of linguistic shifts.</p><p><strong>Results: </strong>AI detection scores increased significantly across all journals (mean: 12.01 to 14.46; + 20.4%), with individual journal increases ranging from +10.0% to +34.5%. The mean number of AI marker words per abstract rose from 0.69 to 0.97 (+41.1%), while high-suspicion abstracts (score ≥ 20) increased from 18.0% to 25.8% (+43%). Changes were consistent across journal specialty, geography, and impact factor, suggesting field-wide adoption.</p><p><strong>Conclusions: </strong>Dental scientific writing has undergone measurable linguistic changes coinciding with widespread AI availability. The consistency and magnitude of these changes raise questions about transparency in AI use, maintaining authorial voice, and preserving the integrity of scientific communication. These findings provide an empirical foundation for evidence-based policies on AI disclosure and appropriate use in academic publishing, grounded in observed changes to scientific communication rather than speculative concern.</p>","PeriodicalId":74682,"journal":{"name":"Research integrity and peer review","volume":" ","pages":""},"PeriodicalIF":9.6,"publicationDate":"2026-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13471417/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148159281","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}