This paper proposes a traceability-based model for author validation of AI-assisted language revision in scientific publications using ChatGPT. Unlike approaches that view the language model primarily as a tool for automated grammatical and stylistic correction, the focus here is on author control, preserving scientific meaning and documenting the changes made. The proposed model comprises sequential components for recording the source text, formulating editorial instructions, classifying the suggested changes, verifying terminological and semantic accuracy, and finally confirming the edited version. Criteria for evaluating AI-edited scientific text are presented, together with an exploratory assessment based on three constructed scientific text excerpts. The main risks associated with excessive automation, changes to terminological nuances, and the blurred line between linguistic assistance and substantive intervention are discussed. The analysis indicates that ChatGPT can be a useful tool for language revision when its use is subject to a clear procedure for author verification, traceability and ethical accountability

ChatGPT,language revision,scientific publications,author validation,traceability,academic writing,artificial intelligence