Generative AI as a conversational companion: A design-oriented post-draft framework for preserving authorial judgment and creative agency in scientific manuscripts

Generative AI is increasingly used in academic writing, yet little guidance exists on how to integrate it without replacing the author’s reasoning or diluting their scholarly voice. Accordingly, the paper is positioned as a design-oriented methodological framework that functions as a metacognitive scaffold by specifying an operational procedure for post-draft AI use rather than as an empirical study or a purely theoretical contribution. In this proposed framework, authors first produce a human-written draft and only then engage generative AI as a diagnostic conversational companion restricted to post-draft auditing of coherence, alignment, and reconstructability. In this revision stage, generative AI is used to perform specific, verifiable actions: checking coherence across sections, auditing alignment between research questions, methods, results, and conclusions, detecting ambiguous or overly abstract formulations, and verifying adherence to reporting conventions. Drawing on Maturana’s concept of structural coupling, we argue that this boundary preserves the author’s epistemic authority by ensuring that the meaning of the manuscript originates in the author’s own distinctions, while AI contributes through perturbations that prompt clarity and refinement rather than generating content. The framework expands current guidelines by moving beyond general ethical prescriptions to define a practical, auditable procedure for post-draft AI use that preserves authorial voice and epistemic control. It offers a replicable workflow that integrates efficiency with transparency, helping researchers use generative AI to strengthen—not substitute—their scholarly reasoning, transforming the role of AI from a content generator into a catalyst for critical thinking and creative precision.