New Research: Saying You Used AI Isn't Enough to Prove You're the Author

Many journals and institutions now require authors to disclose when they use generative AI. But a new article argues that these rules miss the point. Disclosure only documents that AI was used—it doesn't show whether the author truly understands and can defend the work. The author introduces the concept of 'authorial warrant': the idea that you must be able to explain and justify your claims without relying on AI outputs as proof.

The article separates two types of AI use. 'Routine delegation' covers tasks like editing or formatting, which don't affect the substance of your judgment. 'Cognitive delegation' happens when AI shapes your interpretations or justifications—essentially making decisions for you. The author warns that cognitive delegation can undermine the basis of authorship, even if you disclose it.

To help researchers, the article turns disclosure categories into a practical tool. It maps six common research functions (like literature review or data analysis) to potential risks and oversight strategies. For each, it suggests questions to ask yourself to spot cognitive delegation and choose appropriate evaluation methods. This way, authors can ensure they remain responsible for their work.

The article contributes to the ethics of information technology by clarifying when AI-supported scholarship stays under responsible human judgment and when it crosses the line. In a world where AI is increasingly used in research and writing, this framework helps authors, reviewers, and institutions uphold integrity and trust in scholarship.