Chao Cai Libraries and School of Information Studies, Purdue University
Source
Journal of Documentation, Vol. 82 No. 7
DOI: 10.1108/JD-03-2026-0132
Abstract
Purpose – This article introduces the CARDLG framework: Collections, Access, Retrieval, Description, Literacy and Governance. The framework analyzes how generative artificial intelligence (AI) and large language models reconfigure recurring library infrastructural dimensions rather than representing wholly unprecedented disruption.
Design/methodology/approach – The analysis draws on library history from manuscript repositories through digital discovery systems, tracing how eachCARDLG dimension evolved across technological eras.It identifies continuities, accelerations and ruptures in AI/large language model (LLM) implementations compared to prior infrastructural shifts such as card catalogs, online public access catalogs and web-based discovery layers.
Findings – AI/LLMs extend familiar challenges including vendor dependence and opaque retrieval algorithms while introducing distinct risks tied to synthetic answer generation, training data governance and output accountability. The framework reveals how AI integration cuts across all six dimensions simultaneously, requiring coordinated decision-making across collections, systems, instruction and governance units.
Originality/value – This article provides a historically informed analytical framework for library practitioners navigating AI adoption.It offers. CARDLG-based design principles, implementation scenarios across discovery, metadata, literacy instruction and vendor negotiation contexts, and stakeholder implications across library roles. The framework enables deliberate, values-aligned decisions rather than reactive adoption driven by vendor timelines.
