Tracking Editorial Footprints: A Process-Oriented Analysis of ChatGPT Reliance in Student Writing

In this study, we propose a process-oriented framework centered on “editorial footprints,” which we define as the observable steps in a writer’s drafting and revision process when using generative AI. Fifteen female undergraduate students completed two writing tasks using ChatGPT: one under a quick, minimal-effort condition and another under a thorough, high-effort condition. Participants edited a shared rough draft in Google Docs, while their entire interactions with ChatGPT were recorded and qualitatively analyzed. Results show that while the final text lengths were similar, students in the thorough condition made significantly more edits and employed a broader range of ChatGPT prompts, producing work with greater depth, logical coherence, and style consistency, which left more editorial footprints throughout the writing process. These findings reveal distinct patterns of engagement, prompting, and revision between the two conditions and demonstrate the limitations of current AI detectors, which overlook the full scope of the writing process. Our discussion emphasizes that detection of AI-generated writing should incorporate analysis of the writer’s interaction histories and revision behaviors with generative AI tools. We further suggest that understanding these process-based indicators is essential not only for distinguishing AI-assisted writing but also for fostering educational practices that encourage meaningful, reflective engagement with AI in writing.