Reconceptualising AI-Augmented Cataloguing in Libraries

Artificial intelligence is being increasingly explored and adopted by libraries across the globe for various operations and services. Library cataloguing is also seeing experimental implementation of these technologies, yet its roles, risks and governance remain conceptually underspecified. This paper explores how AI can augment cataloguing practice, not merely from the lens of technological determinism, but by duly considering its epistemic and ethical foundations as well. Based on recent implementation of AI technologies such as generative models, machine-learning classifiers, and recommendation systems in metadata workflows, this work first synthesises current practice and identifies critical gaps in transparency, bias management, and professional oversight. It then develops a five-part typology of roles that AI can augment cataloguing, namely: descriptor, classifier, authority resolver, integrator, and recommender, arguing that each of these roles reconfigures longstanding bibliographic control functions in distinct ways. Building on this typology, a four-layered conceptual model is proposed, where these layers represent AI capabilities, Cataloguing practice and its processes, moderating variables including organisational and policy factors and the observable outcomes such as efficiency, accuracy, metadata quality and user retrieval experience. From this model, we articulate a set of hypotheses and propositions to guide future empirical research on AI-augmented cataloguing practice. The paper concludes by suggesting practical guidelines and policy recommendations to ensure that AI strengthens, rather than weakens, the practice of cataloguing by upholding the principles of access, equity and intellectual freedom in knowledge systems.