Artificial intelligence-assisted scientific figure design in international publishing and its implications for editorial responsibility

Abstract

Reviewers, editors, and readers judge a manuscript partly through its figures, which carry evidential as well as rhetorical weight. This review treats figure production as a design problem and as a question of publication equity. Where communicative polish influences how otherwise comparable work is assessed, uneven access to professional illustration, specialized software, mentoring, and visual-literacy support becomes a justice-relevant disadvantage. Kept under human control, artificial intelligence (AI)-assisted tools may narrow part of that gap by helping with figure planning, layout, captioning, accessibility checks, and reproducible plotting. Generative systems also carry their own hazards: fabricated elements, semantic drift, bias, exposure of confidential information, and broken provenance. We propose a five-stage, theory-informed framework in which authors define the scientific claim, choose the visual grammar, draft with bounded AI support, validate every visible element against source material, and prepare files with provenance and disclosure. A worked application to a protein X-ray crystallography workflow shows how the framework can be implemented in a technically demanding setting; the application is explicitly non-empirical. Throughout, AI serves as a drafting and checking aid, not as a source of evidence, and accountability stays with the author.