From resource provider to process auditor: redefining reference roles and workflows in the generative AI era

Purpose This study addresses the disruption caused by generative artificial intelligence (GenAI) in academic libraries, specifically the proliferation of AI-generated hallucinations in student research. It aims to demonstrate how subject services can be restructured from traditional outcome-based evaluation to a process-oriented workflow. By applying change management frameworks and technical services methodologies, the article proposes a strategy to maintain service quality and academic integrity when users employ tools that libraries do not control, shifting the library’s role from resource provision to process assurance. Design/methodology/approach Adopting a case study design at a research institution, this research applies Kotter’s eight-stage change model to redesign support for an interdisciplinary competition. The study employs a mixed-methods approach, comparing data from the 2021 traditional service baseline and the 2025 process intervention cycle, while drawing on the 2023 GenAI-related crisis as the impetus for redesign. Data sources include systematic citation verification audits of over 1,200 citations, pre-post assessments of student AI literacy, service utilization metrics and librarian time-tracking logs. The new process intervention model introduces mandatory research logs and intermediate checkpoints to embed quality control throughout the research cycle. Findings The process intervention model yielded an 82.5% reduction in fabricated citations and a 108% improvement in students’ ability to identify AI hallucinations. While the workflow successfully mitigated AI risks, it required a 689% increase in librarian time, raising scalability concerns. The study confirms that transparency-based approaches requiring research logs are more effective than prohibition, as students who disclosed AI use and followed verification protocols produced fewer errors than those claiming no AI use. Originality/value An underexplored area in the literature is addressed by examining patron-initiated AI use through workflow redesign rather than technology implementation. The contribution lies in applying technical services methodologies, including quality control and standardization, to challenges in subject services. Unlike studies focusing on banning AI or teaching prompt engineering, this research offers an empirically tested process assurance model, redefining the library’s role in the GenAI era from gatekeeper to quality control auditor.