Leading the Transition from AI Disclosure to AI Integrity
.jpeg?width=45&name=Image%20(5).jpeg){kind=small}
Published Date: 20 August 2026 4:40 p.m.
Generative artificial intelligence is rapidly transforming scientific research, raising critical questions about how journals safeguard research integrity. A study published in the MIT Science Policy Review audited AI policies across global medical journals and called for a major shift—moving from static rules that merely police authors to an active, ongoing stewardship of the scientific record.
In this independent evaluation, JMIR Publications was recognized alongside JAMA and PLOS for operating at an AI Governance Readiness Level of 2, the highest level achieved amongst the journals evaluated in the study sample.
| Key Takeaways |
| Top-Tier Governance: JMIR Publications achieved AGRL Level 2—the highest rating in a global audit, alongside JAMA and PLOS—for transparent AI oversight. |
| Rigorous Integrity Checkpoints: The publisher mandates detailed author AI disclosures (including prompts and transcripts) while enforcing strict confidentiality rules for peer reviewers. |
| Proactive Scientific Stewardship: JMIR is pushing beyond static rules toward dynamic governance that continuously evaluates AI tools for bias, validity, and data provenance. |
Key Concepts from the Study
The study evaluates journal oversight using two main frameworks:
-
The PRAIDE Framework: Maps AI integrity checkpoints across six stages of the publication workflow from initial manuscript preparation through peer review and post-publication evaluation.
- AI Governance Readiness Levels (AGRL): Measures a publisher's practical capacity to govern AI risks across five levels:
How JMIR Publications Stacks Up
While most evaluated journals clustered at AGRL Level 1, JMIR Publications achieved AGRL Level 2 (Transparent), demonstrating JMIR Publications’ leadership as an open-access policy diffuser in several key areas:
-
Detailed Disclosures: Strongly recommends authors to provide transparent documentation of AI use, including prompts and interaction transcripts where applicable.
-
Clear Reviewer Guidelines: Establishes explicit boundaries for how peer reviewers may use AI while protecting manuscript confidentiality.
-
Active Enforcement: Maintains clear policies across text, figures, and reviewer accountability.
While we are proud of this recognition, the study rightly notes that static policies across the industry are still lagging behind technological change. Reaching the next stage of governance—where AI workflows are actively evaluated for computational validity, bias, and data provenance—requires continuous learning and shared effort across the global research ecosystem.
Reaffirming Our Commitment to Active Stewardship
JMIR Publications is committed to proactive stewardship, ensuring our practices evolve alongside emerging trends in publishing, shifting scientific norms, and critical integrity demands. Leveraging specialized licensed tools and fostering continuous innovation within our publishing workflows are just a couple of ways that JMIR Publications strives to meet these needs and standards.
“Scientific integrity is a shared responsibility," adds Dr. Tiffany Leung. "We are dedicated to continuously adapting and updating policies to align with technologies, governance recommendations, and scientific norms."
We invite our authors, reviewers, and readers to explore our AI guidelines and join us in building a more transparent future for scholarly publishing.