A sentence-level corpus of explicit affirmative generative AI use disclosures in PubMed Central biomedical articles

It contains 453 manually validated sentence-level records in which biomedical article authors explicitly report using generative AI tools for research or manuscript preparation. Each record includes the disclosure sentence, context, article section, AI tool, usage task, bibliographic metadata, and licensing information.The corpus was created through high-recall retrieval, LLM-assisted classification and extraction, rule-based normalization, and human validation. It can support research on AI disclosure practices and the development of natural language processing methods for disclosure detection and information extraction.