The DOI matches its Crossref record; 24 of the 24 references it cites were found in a registry or an index; 24 of the 64 sources its claims cite were checked for support.
Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment
Open accessAbstract
Fraudulent, falsified, and otherwise unreliable clinical research threatens the foundations of evidence-based medicine because systematic reviews, meta-analyses, and clinical practice guidelines depend on the trustworthiness of the studies they include. Retractions are increasing faster than publication output and often occur only after substantial delay, allowing problematic trials to be cited, pooled, and incorporated into downstream clinical recommendations. Recent evidence shows that retracted randomized trials have contaminated thousands of meta-analyses and hundreds of guideline documents, while only a small minority of affected reviews later self-correct. This is not only a clinical problem but also an integrity analytics problem: unreliable studies generate article-level, trial-level, and network-level signals that are not yet systematically integrated into evidence-synthesis workflows. Existing approaches can detect some image anomalies, textual irregularities, reporting inconsistencies, and statistical red flags, but important blind spots remain. Fabricated clinical datasets may appear statistically plausible, outcome switching often requires registration-publication comparison, and most available tools operate in isolation rather than as coordinated screening systems. We argue that the next generation of safeguards should combine AI/ML-based detection systems, statistical forensic methods, structured trustworthiness appraisal tools, and workflow-level screening frameworks within evidence synthesis. Embedding staged trustworthiness assessment early in systematic reviews, alongside stronger publisher- and database-level integrity infrastructure, could help prevent unreliable trials from distorting pooled estimates and downstream guidance. Such systems should support triage and prioritization rather than replace human judgment. Framed in this way, scalable trustworthiness assessment represents a practical research metrics and analytics agenda for strengthening the evidence ecosystem from primary studies to reviews, guidelines, and policy decisions.
Overview
CoChat Summary
## TL;DR
This paper argues that AI/ML, statistical forensics, and structured trustworthiness tools should be embedded into systematic review workflows to detect unreliable clinical trials earlier and at scale. The goal is to prevent fabricated, falsified, or otherwise problematic studies from distorting meta-analyses, guidelines, and patient care.
## Key findings
- Retractions are rising faster than publication output: the paper cites a 22.09% compound annual growth rate for retracted papers from 2000 to 2022, versus 6.25% for total publications.
- Retractions can materially alter evidence syntheses: the 2025 VITALITY study identified 1,330 retracted randomized controlled trials contaminating 4,095 meta-analyses from 847 systematic reviews; retracted trials changed pooled effect direction in 8.4% of meta-analyses and statistical significance in 16.0% of contaminated meta-analyses.
- Retractions propagate into downstream guidance: evidence from 68 distorted systematic reviews fed into 157 clinical practice guideline documents, and only 5% of systematic reviews corrected themselves after retraction.
- Existing integrity checks catch some problems but miss others: for example, a 2021 audit of 153 clinical trials submitted to Anaesthesia found 44% contained false data and 26% contained fabricated data, while the paper argues that plausible fabricated datasets and outcome switching often evade current tools.
## Methods
- Narrative perspective/review article, not an empirical study.
- Synthesizes previously published evidence, including retraction statistics, audits, and tool descriptions, to argue for a layered AI-assisted trustworthiness workflow in evidence synthesis.
- Uses illustrative case examples and reported study numbers from prior work to motivate proposed screening stages (study identification, eligibility assessment, pre-extraction trustworthiness checks, and targeted forensic review).
## Limitations
- Not stated in the paper.
- Evident from the text: this is a conceptual argument and roadmap, so it does not present a new dataset, experiment, or direct evaluation of the proposed integrated AI system.
- Evident from the text: the paper itself notes that no widely adopted integrated system exists today, so the proposal remains unvalidated at scale.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
23 Valid · 8 Not valid · 24 Not checked · 14 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“In a separate journal-level analysis, identified duplications prompted dozens of editorial actions, including 41 corrections and 5 retractions”
§ Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment > Retractions are rising faster than publication output
What the sources say
- Partly supportedcites Bik et al., 2018 sourceExperimental — accuracy not yet measured
The Results report that inappropriate image duplications in 59 Molecular and Cellular Biology papers led to editorial actions, including 5 retractions. They report 42 corrections, however, not the 41 stated in the claim; no page number is provided in our copy. (LLM, conf=0.91)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“the mean time from publication to retraction is approximately 34 months”
§ Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment > Retractions are rising faster than publication output
What the sources say
- Supportedcites Steen et al., 2013 source
The Results report a mean time-to-retraction of 32.91 months overall for the articles studied, which supports the citing sentence’s approximate figure of 34 months. (LLM, conf=0.93)
- Not matchedcites Lyu et al., 2025
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“often occur only after substantial delay”
§ Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment > Abstract Fraudulent, falsified, and otherwise unreliable clinical research threatens the foundations of evidence-based medicine because systematic reviews, meta-analyses, and clinical practice guidelines depend on the trustworthiness of the studies they include. Retractions are increasing faster than publication output and often occur only after substantial delay, allowing problematic trials to be cited, pooled, and incorporated into downstream clinical recommendations. Recent evidence shows that retracted randomized trials have contaminated thousands of meta-analyses and hundreds of guideline documents, while only a small minority of affected reviews later self-correct. This is not only a clinical problem but also an integrity analytics problem: unreliable studies generate article-level, trial-level, and network-level signals that are not yet systematically integrated into evidence-synthesis workflows. Existing approaches can detect some image anomalies, textual irregularities, reporting inconsistencies, and statistical red flags, but important blind spots remain. Fabricated clinical datasets may appear statistically plausible, outcome switching often requires registration-publication comparison, and most available tools operate in isolation rather than as coordinated screening systems. We argue that the next generation of safeguards should combine AI/ML-based detection systems, statistical forensic methods, structured trustworthiness appraisal tools, and workflow-level screening frameworks within evidence synthesis. Embedding staged trustworthiness assessment early in systematic reviews, alongside stronger publisher- and database-level integrity infrastructure, could help prevent unreliable trials from distorting pooled estimates and downstream guidance. Such systems should support triage and prioritization rather than replace human judgment. Framed in this way, scalable trustworthiness assessment represents a practical research metrics and analytics agenda for strengthening the evidence ecosystem from primary studies to reviews, guidelines, and policy decisions.
What the sources say
- Supportedcites Steen et al., 2013 source
The Results report that retraction took an average of 49.82 months for articles published between 1973 and 2002 and 23.82 months for those published after 2002. They also state that approximately half of articles requiring retraction were retracted within 20 months. These findings support the claim that retractions often occur only after substantial delay; page numbers are not provided in our copy. (LLM, conf=0.92)
- Not matchedcites Lyu et al., 2025
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“for papers authored by serial offenders (“super-retractors”), time to retraction may exceed a decade”
§ Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment > Retractions are rising faster than publication output
What the sources say
- Supportedcites Steen et al., 2013 source
The Results report that articles by Boldt and his associates, described as a serial-fraud incident, took up to 149 months—more than 12 years—to retract. This supports the claim that time to retraction may exceed a decade for papers by serial offenders. (LLM, conf=0.94)
- Not matchedcites Lyu et al., 2025
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedgedstates a possible cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“outcome switching often requires registration-publication comparison”
§ Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment > Abstract Fraudulent, falsified, and otherwise unreliable clinical research threatens the foundations of evidence-based medicine because systematic reviews, meta-analyses, and clinical practice guidelines depend on the trustworthiness of the studies they include. Retractions are increasing faster than publication output and often occur only after substantial delay, allowing problematic trials to be cited, pooled, and incorporated into downstream clinical recommendations. Recent evidence shows that retracted randomized trials have contaminated thousands of meta-analyses and hundreds of guideline documents, while only a small minority of affected reviews later self-correct. This is not only a clinical problem but also an integrity analytics problem: unreliable studies generate article-level, trial-level, and network-level signals that are not yet systematically integrated into evidence-synthesis workflows. Existing approaches can detect some image anomalies, textual irregularities, reporting inconsistencies, and statistical red flags, but important blind spots remain. Fabricated clinical datasets may appear statistically plausible, outcome switching often requires registration-publication comparison, and most available tools operate in isolation rather than as coordinated screening systems. We argue that the next generation of safeguards should combine AI/ML-based detection systems, statistical forensic methods, structured trustworthiness appraisal tools, and workflow-level screening frameworks within evidence synthesis. Embedding staged trustworthiness assessment early in systematic reviews, alongside stronger publisher- and database-level integrity infrastructure, could help prevent unreliable trials from distorting pooled estimates and downstream guidance. Such systems should support triage and prioritization rather than replace human judgment. Framed in this way, scalable trustworthiness assessment represents a practical research metrics and analytics agenda for strengthening the evidence ecosystem from primary studies to reviews, guidelines, and policy decisions.
What the sources say
- Supportedcites Goldacre et al., 2019 source
The Methods (Coding of outcome reporting; page unavailable) describe checking trial publications against outcomes specified before trial commencement, using a dated registry entry when a suitable protocol was unavailable. The Discussion (Policy implications for registries; page unavailable) says registries are often the only publicly accessible time-stamped source of those outcomes. Together, these support the claim that detecting outcome switching often requires comparing registration with publication. (LLM, conf=0.92)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
31 more claims · 7 with their cited sources checked
Create a free account to see all 69 claims and every source verdict.
- It may state two separate claims as one. 27
- Probably not a substantial claim. 13
Citations
Cites
Misconduct accounts for the majority of retracted scientific publications
Fang, F. C., Steen, R. G., & Casadevall, A. (2012) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1212247109 · Cited by 1,232
Why has the number of scientific retractions increased?
Steen, R. G., Casadevall, A., & Fang, F. C. (2013) · PLoS ONE · doi:10.1371/journal.pone.0068397 · Cited by 415
More than 10,000 research papers were retracted in 2023 — a new record
Van Noorden, R. (2023) · Nature · doi:10.1038/d41586-023-03974-8 · Cited by 351
The Prevalence of Inappropriate Image Duplication in Biomedical Research Publications
Bik, E. M., Casadevall, A., & Fang, F. C. (2016) · mBio · doi:10.1128/mbio.00809-16 · Cited by 287 · cited 1×
“Bik and colleagues manually screened over 20,000 biomedical papers and found inappropriate image duplication in ∼ 4%, with at least half of the problematic cases suggestive of deliberate manipulation (Bik et al., 2016).”
COMPare: a prospective cohort study correcting and monitoring 58 misreported trials in real time
Goldacre, B., Drysdale, H., Dale, A., Milosevic, I., Slade, E., Hartley, P., Marston, C., Powell-Smith, A., Heneghan, C., & Mahtani, K. R. (2019) · Trials · doi:10.1186/s13063-019-3173-2 · Cited by 233 · cited 2×
“The COMPare Trials Project, examining 67 trials in five top medical journals, found that each trial reported only 25–96% of pre-specified primary outcomes while silently adding an average of 5.4 new outcomes per trial (Goldacre et al., 2019).”
InfluentialCarlisle, J. B. (2017) · Anaesthesia · doi:10.1111/anae.13938 · Cited by 229 · cited 1×
“…analysis of 5,087 RCTs, a stringent probability threshold flagged ∼ 1.6% of unretracted trials as having highly improbable baseline patterns and, using an author-level heuristic, highlighted Boldt, Fujii, and Reuben, three authors with multiple retracted trials (Mascha et al., 2017; Carlisle, 2017).”
InfluentialFalse individual patient data and zombie randomised controlled trials submitted to Anaesthesia
Carlisle, J. B. (2021) · Anaesthesia · doi:10.1111/anae.15263 · Cited by 173 · cited 2×
“A 2021 audit by editors of the journal Anaesthesia analyzed individual participant data from 153 clinical trials submitted between 2017 and 2020, finding that 44% contained false data and 26% contained fabricated data ( Carlisle, 2021).”
Brown, N. J. L., & Heathers, J. A. J. (2017) · Social Psychological and Personality Science · doi:10.1177/1948550616673876 · Cited by 127
How big is science’s fake-paper problem?
Van Noorden, R. (2023) · Nature · doi:10.1038/d41586-023-03464-x · Cited by 114
Weibel, S., Popp, M., Reis, S., Skoetz, N., Garner, P., & Sydenham, E. (2023) · Research Synthesis Methods · doi:10.1002/jrsm.1599 · Cited by 93
Research works printed but not linked: 5
- Science-wide mapping and ranking of institutions based on affiliated authors Ioannidis J. P. A. Baas J. Boverhof R. Voyant C. bioRxiv [Preprint]
- Retracted randomized trials attributed to super-retractors and top-cited scientists with multiple retractions: secondary analysis of the VITALITY retrospective cohort Lyu C. Matbouriahi M. Naudet F. Ioannidis J. P. A. Cristea I. A. medRxiv [Preprint]. 2025
- AI in Research Integrity: Springer Nature's Innovative Tools Geppetto and SnappShot 2026
- Springer Nature unveils two new AI tools to protect research integrity 2026
- Prevalence and trends in global retractions explored through a topic lens Zhou Z. Lou Y. Shen Z. Li M. arXiv [Preprint] 2025
Metadata
- Authors
- Mo, M., Arora, R., Cao, C., Tricco, A. C., & Moher, D.
- Affiliations
- University of Calgary (CA)
- Harvard University (US)
- University of Toronto (CA)
- St. Michael's Hospital (CA)
- University of Ottawa (CA)
- Corresponding author
- Mo, Moses
- Venue
- Frontiers in Research Metrics and Analytics
- Publisher
- Frontiers Media SA
- Published
- 2026-07-08
- Volume
- 11
- Pages
- 1833223-1833223
- ISSN
- 2504-0537 (ISSN-L)
- PMC
- PMC13388878
- OpenAlex
- W7167700499
- Semantic Scholar
- 6a31c487395ea79500f1d8bbf71aa59b6333080e
- Cited by
- 0
- References
- 24
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Research Metrics and Analytics (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
- Topics
- Social Sciences › Education › Academic integrity and plagiarism
- Social Sciences › Decision Sciences › Statistics, Probability and Uncertainty › Meta-analysis and systematic reviews
- Social Sciences › Library and Information Sciences › Academic Publishing and Open Access
- Keywords
- research integrity, trustworthiness assessment, evidence synthesis, systematic reviews, meta-analysis, AI-based detection, research metrics, publication integrity
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-01 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment. https://app.cochat.ai/papers/10.3389/frma.2026.1833223
- The paper
- Mo, M., Arora, R., Cao, C., Tricco, A. C., & Moher, D. (2026). Artificial intelligence to enhance research integrity in evidence-based medicine: toward scalable trustworthiness assessment. Frontiers in Research Metrics and Analytics. https://doi.org/10.3389/frma.2026.1833223
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.