Transparency, accountability, and scholarly integrity: A systematic narrative review of artificial intelligence governance frameworks in healthcare publication
- 1School of Nursing, Brock University, St Catharines, Canada.
- 2Nursing Department, Federal University of Santa Catarina, Florianopolis, Brazil.
- 3School of Nursing, University of Central Florida, Orlando, Florida, United States of America.
Abstract
The integration of artificial intelligence (AI) in healthcare research and scholarly publication has produced a rapidly expanding but fragmented ecosystem of reporting guidelines, governance frameworks, editorial policies, and regulatory instruments, imposing significant burdens on researchers, reviewers, and editors. This systematic narrative review identified, stratified, and synthesized governance frameworks and reporting standards addressing transparent and ethical AI use in healthcare publications, and characterized the governance trajectory from predictive model standards to generative AI oversight. Fourteen Boolean search strings were executed in PubMed (MEDLINE) covering 1 January 2018–19 May 2026, supplemented by manual searching of the EQUATOR Network guideline library. Records were deduplicated and prioritized using a relevance-scoring algorithm. Of 30,276 deduplicated records, 888 progressed to human screening; 100 publications were included, comprising 69 governance instruments and policies classified across four evidence tiers and 31 contextual evidence sources. Included sources spanned six thematic groups: predictive and diagnostic AI frameworks (G1); sociotechnical and lifecycle governance (G2); generative AI and LLM-specific standards (G3); specialty-specific extensions (G4); surgical reporting updates (G5); and editorial and publication policy (G6). Analysis identified four consecutive phases of governance development from 2018 to 2026, characterized by a shift from model performance transparency toward comprehensive sociotechnical accountability. Three governance tensions emerged: the trade-off between framework specificity and universal applicability; the gap between disclosure-focused standards and lifecycle governance; and the risk of epistemic conflation, whereby heterogeneous instruments are perceived as carrying equivalent authority when treated as interchangeable. AI governance in healthcare publications remains fragmented and epistemically uneven. Advancing the field requires harmonization across evidence tiers, adaptive updating mechanisms, and lifecycle-oriented governance frameworks. Author summary: Artificial intelligence is increasingly used in healthcare research, from developing clinical tools to helping prepare scientific papers. At the same time, many guidelines and policies have been created to tell researchers, journals, and reviewers how this work should be reported and managed. We reviewed this rapidly growing landscape to understand what guidance exists, how it has changed over time, and where important gaps remain. We identified 100 relevant publications, including 69 guidelines, frameworks, and institutional or editorial policies. We found that guidance has evolved from a strong focus on how artificial intelligence models perform toward broader concerns such as fairness, transparency, human oversight, and the use of generative artificial intelligence in research and publishing. However, the current landscape remains fragmented. Different documents vary greatly in how they were developed and how much authority they carry, while many focus mainly on disclosure when a paper is published rather than responsibilities throughout the life of an artificial intelligence system. Our findings suggest that the field now needs better coordination of existing guidance, clearer direction on which standards should be used in different situations, and stronger attention to responsibilities beyond publication.
Related
- Patient expectations for artificial intelligence transparency and agency: A qualitative study and proportionate framework for clinical artificial intelligence implementation
- Early Detection of Red Palm Weevil Infestation: A Systematic Review of Sensing Technologies, Artificial Intelligence, and Field Readiness
- Advances in Artificial Intelligence-Enabled Crop Pest and Disease Detection: A Systematic Review
- Remote sensing and artificial intelligence for crop pest and disease monitoring: A systematic review of sensing technologies, AI approaches, and barriers to field adoption
- Artificial intelligence for detecting autism spectrum disorder: a systematic review and meta-analysis
BibTeX
@article{silva2026transparency,
title = {{Transparency, accountability, and scholarly integrity: A systematic narrative review of artificial intelligence governance frameworks in healthcare publication}},
author = {Amina Silva and Bruna Canever and Monica Lino and Vanessa Silva E Silva},
journal = {PLOS Digital Health},
year = {2026},
month = oct,
doi = {10.1371/journal.pdig.0001783},
url = {https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001783},
note = {CC BY},
}
RIS
TY - JOUR
TI - Transparency, accountability, and scholarly integrity: A systematic narrative review of artificial intelligence governance frameworks in healthcare publication
AU - Amina Silva
AU - Bruna Canever
AU - Monica Lino
AU - Vanessa Silva E Silva
PY - 2026
DA - 2026/10/20
JO - PLOS Digital Health
DO - 10.1371/journal.pdig.0001783
UR - https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001783
AB - The integration of artificial intelligence (AI) in healthcare research and scholarly publication has produced a rapidly expanding but fragmented ecosystem of reporting guidelines, governance frameworks, editorial policies, and regulatory instruments, imposing significant burdens on researchers, reviewers, and editors. This systematic narrative review identified, stratified, and synthesized governance frameworks and reporting standards addressing transparent and ethical AI use in healthcare publications, and characterized the governance trajectory from predictive model standards to generative AI oversight. Fourteen Boolean search strings were executed in PubMed (MEDLINE) covering 1 January 2018–19 May 2026, supplemented by manual searching of the EQUATOR Network guideline library. Records were deduplicated and prioritized using a relevance-scoring algorithm. Of 30,276 deduplicated records, 888 progressed to human screening; 100 publications were included, comprising 69 governance instruments and policies classified across four evidence tiers and 31 contextual evidence sources. Included sources spanned six thematic groups: predictive and diagnostic AI frameworks (G1); sociotechnical and lifecycle governance (G2); generative AI and LLM-specific standards (G3); specialty-specific extensions (G4); surgical reporting updates (G5); and editorial and publication policy (G6). Analysis identified four consecutive phases of governance development from 2018 to 2026, characterized by a shift from model performance transparency toward comprehensive sociotechnical accountability. Three governance tensions emerged: the trade-off between framework specificity and universal applicability; the gap between disclosure-focused standards and lifecycle governance; and the risk of epistemic conflation, whereby heterogeneous instruments are perceived as carrying equivalent authority when treated as interchangeable. AI governance in healthcare publications remains fragmented and epistemically uneven. Advancing the field requires harmonization across evidence tiers, adaptive updating mechanisms, and lifecycle-oriented governance frameworks. Author summary: Artificial intelligence is increasingly used in healthcare research, from developing clinical tools to helping prepare scientific papers. At the same time, many guidelines and policies have been created to tell researchers, journals, and reviewers how this work should be reported and managed. We reviewed this rapidly growing landscape to understand what guidance exists, how it has changed over time, and where important gaps remain. We identified 100 relevant publications, including 69 guidelines, frameworks, and institutional or editorial policies. We found that guidance has evolved from a strong focus on how artificial intelligence models perform toward broader concerns such as fairness, transparency, human oversight, and the use of generative artificial intelligence in research and publishing. However, the current landscape remains fragmented. Different documents vary greatly in how they were developed and how much authority they carry, while many focus mainly on disclosure when a paper is published rather than responsibilities throughout the life of an artificial intelligence system. Our findings suggest that the field now needs better coordination of existing guidance, clearer direction on which standards should be used in different situations, and stronger attention to responsibilities beyond publication.
SN - 2767-3170
ER -