Generative artificial intelligence (GenAI) models such as ChatGPT are increasingly used in academic contexts, yet concerns persist regarding the accuracy of their bibliographic outputs. With the release of GPT-5, OpenAI claims improved factual grounding and reduced hallucination. This study aimed to assess the accuracy, completeness, and error patterns of bibliographic entries generated by ChatGPT-5 within librarianship literature. Employing an evaluative research design, 200 bibliographic entries were generated using GPT-5 and systematically cross-verified against authoritative sources, including Google Scholar and publishers’ websites. Entries were categorized as correct, partially correct (DOI errors only), or incorrect (multiple metadata errors). Quantitative analysis using descriptive statistics and chi-square tests was complemented by qualitative categorization of error trends. Findings revealed that 74% of entries were fully accurate, 20.5% had incorrect DOIs, and 4% contained multiple metadata errors. While core metadata elements such as author, title, year, and journal were consistently present, critical details such as volume, issue, page numbers, and valid DOIs were often incomplete or inaccurate. Statistical results confirmed significant associations between metadata completeness, DOI presence, and bibliographic accuracy. The study concludes that GPT-5 shows meaningful improvement over earlier versions but cannot yet replace human verification in bibliographic work.

Recent advances in machine learning and artificial intelligence have provided more alternatives for the implementation of repetitive or monotonous tasks. However, the development of AI tools has not been straightforward, and use case exploration and workflow integration are still ongoing challenges. In this work, we present a detailed qualitative analysis of the performance and user experience of popular commercial chatbots when used for document classification with limited data. We report the results for a real-world example of metadata augmentation in academic libraries environment. We compare the results of AI chatbots with other machine learning and natural language processing methods such as XGBoost and BERT-based fine tuning, and share insights from our experience. We found that AI chatbots perform similarly among them while outperforming the machine learning methods we tested, showing their advantage when the method relies on local data for training. We also found that while working with chatbots is easier than with code, getting useful results from them still represents a challenge for the user. Furthermore, we encountered alarming conceptual errors in the output of some chatbots, such as not being able to count the number of lines of our inputs and explaining the mistake as “human error.” Although this is not complete evidence that AI chatbots can be effectively used for metadata classification, we believe that the information provided in this work can be useful to librarians and data curators in developing pathways for the integration and use of AI tools for data curation or metadata augmentation tasks.

Academic libraries in Ghana have made progress in the adoption of integrated library management systems. This study sought to find out the application of systems and standards in cataloguing information resources in Ghanian Academic libraries. Interviews and questionnaires were used to solicit response from four academic libraries in Ghana. The study identifies lack of implementation of the current international cataloguing standard; Resource Description and Access (RDA) and also low level of knowledge and use of BIBFRAME and Dublin Core metadata standards.

Changes to terminology take time and heighten tensions in language description, in preference of naming conventions, and institutional practices. External forces like the mandates of the United Nations Declaration on Indigenous Peoples, and the Canadian Federation of Library Associations (CFLA) recommendations for the Truth and Reconciliation Calls to Action, influence the actions taken by organizations and move us forward. However, other structural and systemic forces can impede these efforts. At the University of Calgary, decisions about which vocabularies to use are further muddied by different practices across our units, and the methods available to make updates to our systems. Our Library Managment System (LMS) needs to wait for updates from the vocabulary authorities, while our digital collections does not, allowing them to make big changes faster. By examining the applicable vocabularies in Canada, we can surface the hegemonic forces at work, that exist internal and external to the institution. For instance, while standardization aids in discovery, it also drives a hegemonic use of language which does not describe Canadian content such as Indigenous names. In grappling with these forces, we confront and oppose them as we work through the process of updating subject headings and descriptive language for Indigenous content within our systems.

The integration of generative Artificial Intelligence (AI) tools such as ChatGPT-4o is redefining traditional workflows in library technical services, particularly in cataloguing and bibliographic functions. This study, therefore, investigates the awareness, use, and perceived challenges of ChatGPT-4o among cataloguers and bibliographers in Botswana, Nigeria, and South Africa. Using a phenomenological research design, six professionals attending the 5th Biennial International Conference of the Department of Library and Information Studies at the University of Botswana were purposively selected. Structured interviews were conducted, and data were analyzed through narrative analysis to uncover participants’ experiences. Findings revealed that all participants were aware of ChatGPT-4o and demonstrated varying degrees of self-taught application in cataloguing and bibliography. The tool was frequently used for MARC record generation, subject indexing, citation formatting, and bibliographic compilation. Participants highlighted substantial benefits, including enhanced speed, reduced workload, and improved thematic coverage. However, notable limitations such as AI hallucinations, lack of contextual sensitivity, internet dependence, and digital literacy gaps were also identified. The significance of this study lies in its empirical exploration of real-world experiences with ChatGPT-4o from cataloguing professionals in an African context, filling a crucial gap in existing literature. The study concludes that ChatGPT-4o holds significant promise as a complementary tool for technical service delivery but must be used with human oversight. It recommends policy development, staff training, and ethical guidelines for responsible integration into library workflows.

This paper reviews existing peer-reviewed literature concerning the application of artificial intelligence (AI) technologies in the context of library cataloging work published since the public release of ChatGPT in 2022. Patterns of analysis in the literature are identified, rigor of investigations assessed, and areas for future work indicated. Existing peer reviewed literature tends to have optimistic-to-positive evaluations of the usefulness and applicability of AI technologies to the creation and maintenance of library metadata, but generally lacks compelling experimental evidence to support its findings. Additional investigation is necessary to establish performance benchmarks for both human and AI-assisted cataloging in order to adequately assess the desirability and efficacy of AI integration into library cataloging.

This article will discuss generative artificial intelligence (GAI) and cataloging/metadata creation in academic libraries, focusing on recent research and recommendations. This article will also explore the following questions: How can cataloging/metadata professionals at academic institutions incorporate general frameworks and standards about ethical AI implementation into policies for AI use in their workflows? How can technical services/cataloging departments promote their skills to provide accurate and culturally sensitive metadata? More importantly, how do cataloging/metadata professionals ethically fill the gap between the cataloging/metadata profession and the new world of AI without sacrificing job security?

The quality of metadata in bibliographic records can be compromised by issues of incompleteness and inaccuracy. One of the most significant challenges is the inaccurate representation of author names. In the specific case of metadata in the MARC21 format, the MARCQuality tool—developed by researchers at the Central University “Marta Abreu” of Las Villas—partially addresses these issues by unifying variations of the same author’s name and resolving, to some extent, synonymy. However, limitations remain concerning homonymy, where multiple authors have the same name. This study aims to implement a solution that optimizes the preparation of MARC21 record fields for future use in neural network models designed for author name disambiguation. As these models are not yet available, the proposed approach focuses on structuring and adapting the fields to improve prediction accuracy. The solution is based on disambiguation techniques using LAGOS AND (Large, Gold Standard Dataset for Scholarly Author Name Disambiguation), as established in HFAND (Hybrid Framework for Author Name Disambiguation). Given the differences between MARC21 cataloged records and LAGOS AND publications, an analysis is necessary to establish correspondences between their fields. This analysis also incorporates the Dublin Core format, used by UCLV’s DSpace repository, to explore its potential integration into the MARCQuality tool. As a result, a structured procedure is developed to organize record fields for preprocessing and similarity metric calculations, facilitating their application in neural network models for author name disambiguation.

The goal of this essay is to critically analyze how metadata and cataloging work have adapted due to advances in digital technology and artificial intelligence (AI). It argues that while these tools can streamline, expedite, and enhance metadata work for library professionals, they also exacerbate hazards inherent in a late capitalist landscape, where the focus on efficiency and profitability can subvert the creation of accurate and nuanced metadata. It is my aim to illuminate the operations of late capitalism on the commodification of library metadata, where the ethical risks outweigh the reward of functionality. This essay serves as a counter viewpoint to the push for increasing AI’s role in metadata generation and calls for librarians to thoughtfully and innovatively engage with AI technologies, while resisting the unchecked commodification of information. By investigating these intersections, this article seeks to provide a pathway for metadata professionals to leverage emerging technologies responsibly, ensuring they can continue to serve as stewards of reliable and accessible information.

The opacity of AI systems leads to challenges related to algorithmic bias, data sovereignty, and regulatory compliance. This study explores the role of metadata and paradata as mechanisms for embedding ethical oversight into AI development. It employs a qualitative approach, including a literature review and conceptual analysis, to examine how these elements contribute to ethical AI oversight. It proposes an ethical AI governance framework structured around five key principles: (1) standardized and dynamic metadata and paradata models, (2) interdisciplinary collaboration, (3) policy and regulatory interventions, (4) capacity building, and (5) a unified framework for metadata and paradata standards. Findings indicate that metadata and paradata enhance AI fairness by ensuring traceability and regulatory compliance. Dynamic models allow real-time updates, improving bias mitigation and accountability. However, challenges such as the lack of standardized documentation, regulatory complexities, and the need for emerging technologies like blockchain must be addressed. Future research should focus on automating metadata and paradata management to improve scalability. By implementing the proposed framework, stakeholders, including AI developers, policymakers, and metadata professionals, can foster responsible AI practices that align with ethical principles, regulatory requirements, and societal values.