Generative AI for Bibliographic Description: What Works, What Doesn’t

Abstract

As interest in applying artificial intelligence (AI) to cataloging and metadata creation grows, there remains a lack of comparative analysis on how current generative AI performs in real-world workflows. This project evaluates the practical capabilities of four freely available generative AI models in creating descriptive metadata, examining which aspects of bibliographic description they handle effectively and where they fall short. By analyzing each model's output and identifying strengths and limitations, the study offers guidance for catalogers seeking to integrate generative AI into their work. The findings aim to support informed decision-making, set realistic expectations, and contribute to ongoing discussions around automation, labor, and quality in metadata creation.