Human-first, AI-assisted Peer Review: A Proposed Workflow for Community Medicine and Public Health Journals.

Abstract

Growth in community medicine and public health publications has increased peer-review workload and the risk of missed methodological and statistical issues. Because peer review is largely voluntary and often performed alongside clinical, teaching, administrative, and research duties, reviews may be delayed or rushed. Generative artificial intelligence (AI) may help, but unstructured use raises concerns about confidentiality, integrity, and automation bias. This viewpoint proposes a human-first, AI-assisted workflow in which reviewers first complete an independent manual review. With editor permission, a constrained AI tool is then used only for checklist-based cross-checking of reporting completeness and internal consistency. Conceptually, this workflow may reduce missed checklist items, improve consistency checks, and support reviewer calibration. Purpose-built, standardized AI tools may offer additional advantages through predefined task boundaries, uniform outputs, and greater control over data retention and disposal. AI-assisted peer review may be acceptable in community medicine and public health journals if journals require permission before use, approved confidentiality-protecting systems, data minimization, disclosure to editors, and full reviewer accountability. The same principles could later be adapted for other disciplines.