Abstract
The use of generative AI (GenAI) to aid in the writing of research articles is potentially highly beneficial, yet the practice runs the risk of producing texts that are not the author's original contribution. Based on distributed cognition, this research focuses on using GenAI prompts of different strengths to revise an article. Eight extracts of early drafts of human-written articles were revised using four prompts ranging from light editing to heavy editing. The resulting texts were evaluated on five criteria. Three criteria showed similar patterns (tagging by an AI detector, proportion of original wording retained, and use of rare words) with the lightest edited texts being around 90% original contribution, and the heaviest edited texts around 20% original contribution. Offloading the processes of revising articles to GenAI can result in loss of authorial ownership. The implications of the findings for journals and for GenAI training for researchers are discussed.
Keywords: Generative AI, research articles, editing, revising, distributed cognition, authorial ownership