Scientific writing in the age of artificial intelligence: trust on trial?

Abstract

The rapid integration of generative artificial intelligence (AI) is transforming scientific writing and publishing, creating both unprecedented opportunities and critical ethical challenges. This article investigates how the use of AI tools affects research integrity, authorship accountability, and peer review processes in scientific publishing. Methodologically, the review synthesizes literature on current AI policies, detection tools, and empirical surveys of author and reviewer practices. Three key hypotheses are proposed for future empirical testing: (H1) mandatory AI disclosure improves the detection of fabricated content; (H2) AI-assisted language refinement enhances manuscript clarity without compromising originality; and (H3) undisclosed AI use by reviewers diminishes the depth of critique. The main findings indicate dominant reliance on descriptive studies, highlighting the need for hypothesis-driven, cross-disciplinary research frameworks and greater transparency to ensure that AI adoption fortifies the trustworthiness of scholarly communication.