Artificial intelligence-assisted manuscript text and visuals: Supporting scientific communication while preserving scientific integrity

Abstract

Artificial intelligence is increasingly used in manuscript preparation, from language refinement and text generation to the creation and modification of scientific visuals. However, responsible use of artificial intelligence in scholarly publishing depends less on whether artificial intelligence was used than on the role it played in producing material submitted for publication. This paper summarises artificial intelligence use across two domains: manuscript text and manuscript visuals. For text, we distinguish between artificial intelligence applications that primarily support communication, such as grammar and language refinement, and those that generate or substantively shape scholarly content. The latter raises greater concerns regarding accuracy, authorship, verification, and disclosure. For visuals, we distinguish between artificial intelligence-assisted preparation of explanatory materials, such as schematic diagrams, and artificial intelligence-generated or artificial intelligence-altered images that purport to represent research data, uses that are subject to substantially greater restrictions under current publishing guidance. Drawing on recommendations from the International Committee of Medical Journal Editors, the Committee on Publication Ethics, and representative publisher policies, we summarise the established principle that human authors remain responsible for the accuracy, interpretation, and integrity of all artificial intelligence-assisted content to different forms of artificial intelligence use in manuscript preparation. We further highlight that expectations regarding disclosure, permitted uses, and restrictions on artificial intelligence-generated or Artificial intelligence-altered content vary across journals and publishers.