GAIDeT (Generative AI Delegation Taxonomy): A taxonomy for humans to delegate tasks to generative artificial intelligence in scientific research and publishing

BACKGROUND: The integration of generative artificial intelligence (GAI) in research raises concerns about transparency, accountability, and task delegation. While frameworks such as CRediT and the NIST AI Use Taxonomy address contributions to research, they either exclude AI-assisted input (CRediT) or do not provide a stage-specific approach (NIST). A structured taxonomy is needed to delineate GAI's contributions across research stages while preserving human oversight and research integrity. METHODS: This study introduces the Generative AI Delegation Taxonomy (GAIDeT), informed by existing contributor role taxonomies, peer-reviewed literature, and an iterative consensus-building approach. It categorizes GAI's contributions at macro and corresponding micro levels, specifying the degree of human oversight required. RESULTS: GAIDeT provides a structured framework for documenting GAI's role in scholarly research. It classifies research activities into key domains - conceptualization, literature review, methodology, data analysis, writing, supervision, and ethical review - ensuring transparency and human accountability. A GitHub-based interactive tool - the GAIDeT Declaration Generator - was developed to help researchers document delegation choices transparently. CONCLUSIONS: By standardizing GAI task delegation, GAIDeT enhances research integrity and transparency. Future work should focus on empirical validation, cross-disciplinary adaptability, and policy implications for GAI governance.