Academic libraries face a fundamental transformation as artificial intelligence (AI) and vendor consolidation reshape professional practice and institutional autonomy. This paper argues that these changes create what it terms a franchise model of librarianship, where library workers operate platforms controlled by vendors rather than exercising professional judgment over knowledge organisation and information access. Like franchise operators who serve the products their franchisors specify according to procedures mandated by headquarters, librarians increasingly administer systems they neither control nor can meaningfully modify, while remaining accountable to users for service quality.
This transformation matters because it fundamentally alters the nature of library work, the quality of information access, and the ability of libraries to fulfil their democratic mission. Discovery systems silently censor research topics (
While the empirical documentation draws primarily from US academic libraries, where
The franchise business model, as established in organisational economics, involves a distinctive structural relationship: a franchisor owns proprietary systems and intellectual property while a franchisee pays ongoing fees for the right to operate using those systems under the franchisor’s terms (
This paper applies the franchise model not as metaphor but as a structured analytical lens. The mapping is precise: vendors function as franchisors who own proprietary platform infrastructure (Alma, WorldShare, Primo) and control development roadmaps, feature implementation, and default configurations. Libraries function as franchisees who pay substantial subscription fees to operate these systems and serve users through interfaces determined by vendors. The Software–as-a-Service (SaaS) delivery model functions as operational standardisation, as libraries receive the same platform with limited capacity for local customisation. When vendor systems fail, libraries absorb the consequences experienced by users as a result of decisions made at corporate headquarters, just as franchise operators absorb customer complaints about products they did not design and cannot alter.
The franchise framework was chosen over alternative conceptualisations—outsourcing, platform dependence, colonisation—because it captures a quality distinctive to the library-vendor relationship that these alternatives miss. Unlike outsourcing, franchise operators still perform the labour under corporate terms, maintaining the appearance of autonomous professional service. Unlike simple platform dependence, the franchise model identifies the specific mechanisms through which autonomy is eroded: fee extraction, standardised operations, externalised risk, and the elimination of professional expertise. Unlike colonisation metaphors, the franchise framework identifies how libraries participate in their own transformation through institutional choices shaped by structural pressures.
This paper introduces the franchise model as an analytical framework for understanding the structural transformation of academic librarianship, extending existing theories of platform labour and surveillance capitalism to institutional settings where professional expertise has historically mediated public access to knowledge. The paper addresses three interconnected research questions: first, how do vendor consolidation and AI implementation structurally transform the professional practice of academic librarianship? Second, in what ways does this transformation create franchise conditions where libraries operate infrastructure controlled by vendors while bearing accountability for service quality they cannot ensure? Third, what forms of resistance emerge within and against this structural transformation, and what do they reveal about alternative futures for library technology governance?
The central thesis is that the convergence of vendor consolidation and AI implementation creates a franchise model of academic librarianship in which libraries retain institutional identity and accountability to users while ceding substantive control over information organisation, discovery, and access to corporate platform operators. This franchise model emerges from the intersection of corporate consolidation with institutional conditions such as declining funding, administrative enthusiasm for AI-driven efficiency, and professional metrics prioritising comprehensiveness that create the structural preconditions for platform dependence.
The transformation this paper analyses manifests concretely in daily library practice. An undergraduate student at a University of Wisconsin system library approached the reference desk with concerns about an article discovered through Ex Libris Primo, connected to the Alma library services platform and drawing from the Central Discovery Index (CDI)—all Clarivate products serving the majority of research libraries. The student had employed the source evaluation practices central to everyday reference instruction: searching the library database, finding a peer-reviewed article, verifying its ISSN and DOI, confirming CrossRef registration, and noting the journal’s indexing alongside established publications. While critical information literacy scholarship has long advocated moving beyond such verification by checklist evaluations toward structural analysis of how information systems shape knowledge access (
What prompted the student’s scepticism was attempting to verify the article’s citations. Library staff discovered that of the article’s thirty-two listed references, only five could be verified as existing publications; the remaining 84% were fabricated. The student’s report, as summarized by the reference services coordinator in the library’s support correspondence, raised urgent questions about the credibility of the institution’s library resources and whether systems supplied by vendors could reliably deliver quality scholarly content (Ex Libris Support, Help Ticket #552419, October 8, 2025).
The journal, published by Al-Kindi Center for Research and Development, carried every institutional marker of academic credibility including DOI registration, ISSN assignment, CrossRef indexing, and discovery system inclusion. Yet closer inspection revealed telling inconsistencies: the corresponding author’s contact was a Gmail address rather than institutional email, the listed departmental affiliation contained errors, the editor-in-chief worked in an unrelated discipline, and the publisher’s listed London address appeared to be a residential dwelling.
When system librarians contacted Ex Libris, the vendor’s response illuminated the franchise dynamic with remarkable clarity. Ex Libris explicitly disclaimed responsibility for quality control, stating that it does not systematically vet or exclude individual journals and instead defers to broader community and industry standards. The vendor indicated that decisions to block or remove content are handled reactively, on a case-by-case basis, typically following reports from the library community rather than through proactive monitoring (Ex Libris Support, Help Ticket #552419, October 8, 2025).
The vendor positioned itself as neutral infrastructure that merely transmitted content, explaining that proactive monitoring was constrained by the lack of authoritative, standardized lists of predatory publishers and the overall complexity of the scholarly publishing landscape. Rather than committing to remediation, Ex Libris recommended that libraries take responsibility for communicating to their users the known limitations and risks of the systems those libraries pay to provide (Ex Libris Support, Help Ticket #552419, October 8, 2025). This captures the franchise model’s central dynamic: libraries pay substantial subscription fees for discovery infrastructure; when that infrastructure fails, the vendor disclaims responsibility and externalises remediation labour to subscriber institutions. After weeks of escalation through multiple support tiers, Ex Libris eventually removed the specific journal, a process that required sustained professional labour to remove a single predatory journal from a database containing millions of records. The vendor’s suggestion that libraries could manage the problem themselves by deactivating the journal on their end if removal on the vendor side proved impossible (Ex Libris Support, Help Ticket #552419, October 8, 2025) further illustrates the dynamic that libraries manage quality within systems they pay for but do not control.
This case illuminates why the franchise model matters beyond labour concerns. The student was practising exactly the kind of critical engagement that information literacy instruction aims to develop, questioning surface-level credibility markers and seeking verification, an impulse aligned with what critical information literacy scholars advocate (
The franchise model draws on three interconnected bodies of scholarship: the political economy of information and platform capitalism, which explains the structural logic driving vendor consolidation; the sociology of professions and critical librarianship, which reveals how platform dependence erodes professional autonomy; and critical information literacy, which documents the consequences of algorithmic mediation for knowledge access.
The transformation of library technology markets reflects broader patterns of information commodification.
The franchise model fundamentally concerns the erosion of professional autonomy.
Critical librarianship scholarship applies these dynamics to library contexts.
Empirical research on AI cataloguing reveals significant limitations. Multiple independent studies, including the Library of Congress’ computational description experiment (
The critical information literacy movement has developed substantial alternatives to source evaluation approaches that rely on checklists.
Together, these literatures establish the theoretical conditions the franchise model synthesises. Platform capitalism provides the extractive logic; professional autonomy erosion provides the mechanism; and compromised information literacy infrastructure provides the consequences for users.
The franchise model does not emerge solely from vendor strategy. It arises at the intersection of corporate consolidation and institutional conditions that create structural preconditions for platform dependence.
Declining library budgets represent the most fundamental precondition. The shift from print collections that libraries owned to electronic resources they licence transferred not only access models but institutional leverage. Libraries that owned collections retained permanent access and could negotiate from material independence, while libraries dependent on licensed content must continuously pay vendors, creating the ongoing fee relationship characteristic of franchise arrangements.
The profession’s historical emphasis on collection size as a quality metric created incentives aligning with vendor interests in comprehensiveness over verification. When discovery systems index billions of items, scale is marketed as a feature. The verification crisis documented in this paper’s case study is, in part, a consequence of collection metrics that prioritise breadth over the professional judgment required to ensure quality.
University administrative culture provides the third enabling condition.
These conditions do not excuse vendor behaviour, but they explain why the franchise model succeeds. Vendors built business models that exploit institutional pressures, offering platform solutions that appear to address funding crises while transferring professional control to corporate infrastructure.
This paper employs a critical qualitative research design positioned within the tradition of critical information studies and the political economy of communication. The analysis synthesises multiple evidence streams to examine how vendor consolidation and AI implementation transform academic librarianship, applying the franchise model as an analytical framework following the methodological tradition of critical analysis in library and information science (
The paper examines five distinct categories of AI technology, recognising that
First, supervised machine learning for subject classification encompasses tools, like Annif, developed by the National Library of Finland, which use statistical models trained on existing cataloguing records to predict subject headings and classification numbers. Second, large language models for cataloguing describes the application of generative AI such as GPT-4 to creating bibliographic records in MARC format, raising distinct concerns about hallucination and the generation of plausible but inaccurate metadata. Third, neural content filtering refers to multi-class classification models, specifically Microsoft Azure OpenAI’s content safety system, which score user prompts for potentially harmful content. When integrated into library discovery tools, these filters determine which scholarly queries receive responses and which are blocked.
Fourth, algorithmic ranking and relevance describes the proprietary algorithms determining order and visibility of search results in discovery systems, shaping information access through relevance determinations that operate largely without transparency. Fifth, AI metadata generated by vendors describes tools such as Clarivate’s auto-enabled AI metadata for Ebook Central, announced to subscribing libraries in October 2025, for which no public disclosure of the specific technical implementation is available (
The analysis draws on six categories of evidence: empirical studies of AI accuracy in cataloguing (
The verification crisis case study was selected as what
Evidence was analysed through the franchise model framework, with each stream examined for how it illustrates structural features of the franchise relationship: fee extraction, operational standardisation, externalised risk, franchisor control of product specifications, and erosion of franchisee expertise. The analysis proceeds inductively from documented evidence to structural interpretation.
The empirical evidence draws primarily on US academic library market data; comparable concentration data for other national contexts would strengthen external validity. The case study represents a single institution; systematic research across multiple institutions would establish prevalence. Access to proprietary vendor data, algorithm specifications, content filtering thresholds, AI training data, remains unavailable, limiting analysis to observable outputs.
The author occupies a dual position as both a practising cataloguing librarian at a university library and a critical researcher examining the systems that structure daily professional work. This positionality provides access to practitioner knowledge and documentary evidence while positioning the analysis within a particular professional perspective. The author has sought to balance insider knowledge with systematic evidence from peer-reviewed scholarship, industry data, and professional documentation, presenting vendor positions through their own communications rather than characterising them solely through a critical lens.
Marshall Breeding’s longitudinal analysis reveals a dramatic transformation of library technology markets. While forty-three vendors competed in 1990, ProQuest controlled 72% of the overall academic library services platform market and 84% at Association of Research Libraries (ARL) member institutions by 2020, with Herfindahl-Hirschman Index scores of 6,934 far exceeding the Federal Trade Commission’s 2,500 threshold for
SPARC’s opposition to the Clarivate-ProQuest merger argued it would bring the market close to monopoly conditions. The merger was expected to generate over $100 million in cost synergies through streamlining facilities, technology infrastructure, and personnel; synergies achieved by eliminating professional expertise while retaining subscription revenue (
When Clarivate announced auto-enabled AI metadata generation for Ebook Central bibliographic records in October 2025, the communication exemplified franchise dynamics. Clarivate notified subscribing libraries that the feature would be activated automatically for all customers, with the option to disable it in advance of the release through system settings, an opt-out rather than opt-in model (
Vendors control development roadmaps, feature implementation, and default settings while libraries manage consequences experienced by users. Librarians cannot modify vendor algorithms but must explain to researchers why searches for
OCLC’s 2025 elimination of approximately eighty positions while developing AI cataloguing tools exemplifies the franchise contradiction. The organisation justified layoffs citing AI’s influence, yet AI systems depend entirely on training data created by the workers being displaced. Northwestern University’s 2025 layoff of 200 employees, including Violet Fox, a leading voice in critical cataloguing who pioneered reparative description and served as one of the editors of the Dewey Decimal Classification, illustrates how institutions eliminate specialised expertise even as they claim to value ethical practice.
The broader workforce trajectory confirms systematic decline: library professionals in the United States numbered 289,400 in 2024, down from 395,000 in 2006—a 26.7% decline (Department for Professional Employees, AFL-CIO, 2025). The American Library Association Core Subject Analysis Committee’s Devaluation of Cataloguing Working Group (2024) captures the franchise trap precisely by placing the burden of justification on the profession itself, even as vendors automatically deploy features intended to displace that expertise.
The Library of Congress Labs’ Exploring Computational Description experiment tested five opensource machine learning models on approximately 23,000 eBooks, measuring performance using the F1 score, a standard classification metric combining precision and recall on a 0-100% scale. Large language models achieved only 26% F1 accuracy for Library of Congress Subject Headings, while Annif reached 35% for text classification (
No other profession would accept comparable failure rates, yet the franchise model explains why libraries do. The cost savings from eliminating cataloguers are immediate and quantifiable, e.g., OCLC reported $225 million revenue for fiscal year ending June 2023 (
The 2025 Library Perceptions survey reveals an unexpected finding: customer satisfaction scores remain moderate to high despite widespread recognition of monopolistic practices. Alma maintains satisfaction scores of 7.02 among large academic libraries, yet respondents express worry about Clarivate’s motivations and describe feeling trapped by vendor lock-in (
As documented in the case study, the discovery of articles with 84% fabricated references represents systemic failure of verification infrastructure. CrossRef accepts DOI registration without verifying reference accuracy; ISSN provides identification without quality control; vendor discovery systems index content based on fees rather than scholarly merit. Vendors have no incentive to exclude questionable content that pays indexing fees and inflates comprehensiveness metrics that libraries themselves use to justify subscription expenditures. The result is an information literacy crisis understood not as pedagogical challenge but as infrastructure failure. Even sophisticated critical approaches face structural limits when the infrastructure has been hollowed out.
The WorldCat paradox exemplifies how platform capitalism transforms collective professional resources into corporately controlled property. Libraries have contributed cataloguing labour to WorldCat since 1971, yet OCLC asserts significant control with contractual restrictions preventing libraries from freely sharing records they created. When Anna’s Archive redistributed WorldCat data in late 2023, OCLC responded by filing suit in January 2024, characterising the scraping as cyberattacks and hacking (
The Clarivate consolidation creates vertical integration across the entire research lifecycle: one entity now controls discovery systems (Primo, Summon), library services platforms (Alma, Sierra, Polaris), research analytics (Web of Science, InCites), citation management (EndNote, RefWorks), and research workflow tools (
Content filtering in AI-enhanced discovery represents the franchise model’s most direct control mechanism. Microsoft’s Azure OpenAI content filtering, which powers Clarivate’s Primo and Summon Research Assistants, employs neural classification models designed for consumer applications rather than scholarly research (
Ex Libris’s response exemplifies the franchise chain: filtering ’is due to safeguard policies enforced by our AI service provider (Microsoft Azure) to support ethical and responsible AI use. It’s not something the Primo or Summon applications control’ (quoted in
The evidence demonstrates systematic transformation of librarianship from professional practice to platform administration. Librarians may no longer primarily create original cataloguing but review AI-generated metadata achieving 26% accuracy; no longer design discovery interfaces but configure vendor platforms that censor research topics; no longer exercise professional judgment but implement auto-enabled vendor decisions. This represents what
Despite structural constraints, resistance practices emerge across multiple domains. The University of California system’s 2019 termination of its Elsevier contract, generating nearly 10% of all US research output, walking away from a contract exceeding $10 million annually, demonstrates that franchise conditions can be challenged through coordinated institutional leverage. While Elsevier terminated direct access to post-2018 content, UC retained perpetual access to approximately 95% of previously licensed materials (
Community-developed systems demonstrate increasing viability. Koha, arguably the most widely deployed ILS globally, has become predominant in Latin America, India, and lower income countries, a significant finding demonstrating that Global South institutions have chosen community–governed alternatives over vendor franchise arrangements (
Alternative information models operating outside vendor platforms demonstrate the continued viability of non-commercial knowledge organisation. Zine libraries and collections, including the Barnard Zine Library, Denver Zine Library, and Queer Zine Archive Project, operate without vendor mediation: no licensing agreements, no usage surveillance, no algorithmic discovery. These collections function through direct human curation and community exchange, preserving alternative models of information stewardship that prioritise community accountability over commercial efficiency.
Professional governance frameworks address AI implementation directly. The PCC Task Group on AI and Machine Learning for Cataloging and Metadata, charged in September 2024, emphasises that ’AI is intended to augment, not replace, the valuable work of cataloguing professionals’ and ’it is not a viable strategy for reduction of personnel’ (
These resistance practices share a critical insight: technical improvements alone cannot address political economy. Improving AI accuracy does not restore professional autonomy when algorithmic systems operate through infrastructure controlled by vendors. The franchise model reveals that these are structural power relationships requiring collective political action, not technical problems requiring technical solutions.
The franchise model extends platform capitalism theory to professional service contexts. While
The case challenges assumptions about AI augmenting human expertise. In library contexts, AI serves primarily to justify labour elimination rather than enhance capabilities—the 26% accuracy rate matters less to vendors than the ability to claim technological modernisation while cutting costs. This suggests AI implementation in professional contexts may systematically prioritise labour displacement over functional improvement when implemented through platform relationships structurally akin to franchises, a finding with implications for other professional domains. The information literacy implications connect the franchise model to critical pedagogy scholarship. Traditional information literacy pedagogy assumes the stable verification markers named above (DOIs, ISSNs, peer review, library endorsement) remain reliable, but the case study demonstrates they are systematically compromised. Even the more sophisticated critical approaches advocated by
The franchise framework suggests strategic priorities: demanding vendor disclosure of AI training data and accuracy rates as contractual requirements; coordinating multi-institutional responses following the UC-Elsevier model; investing in open-source infrastructure as demonstrated by FOLIO and ReShare; establishing enforceable quality thresholds through professional governance bodies; developing algorithmic literacy instruction that teaches students how discovery systems reflect business models; and systematically auditing content filtering as a condition of continued subscription. These responses require recognising that technical problems reflect political economy; the franchise model’s analytical value lies in directing attention from technical fixes to the power relationships determining how technology is implemented.
While the empirical documentation draws primarily from US academic libraries, the structural dynamics operate globally. Systematic comparative research would reveal whether franchise dynamics manifest similarly under different regulatory frameworks (GDPR in Europe, varying data sovereignty laws), in different library sectors (public libraries, Global South institutions), or at different levels of vendor consolidation. The franchise model provides a testable analytical framework for such comparisons. Future research should also examine long-term impacts of AI metadata on research outcomes, user behaviour adaptations to algorithmic mediation, effectiveness of algorithmic literacy instruction, and economic models for sustainable alternative infrastructure.
The franchise model reveals how AI implementation and vendor consolidation transform academic librarianship under digital capitalism. Like franchise operators, librarians administer systems they do not control, implement features they did not request, and remain accountable for quality they cannot ensure. The framework identifies specific structural mechanisms — fee extraction, operational standardisation, externalised risk, jurisdictional displacement — and connects them to the institutional conditions that sustain vendor dependence.
Yet the evidence also demonstrates that the franchise model does not describe an inevitable future. The University of California’s challenge to Elsevier, Australia’s adoption of open-source resource sharing, FOLIO’s emergence as a viable alternative, Koha’s global deployment, PCC governance frameworks, and union organising all indicate that collective action can contest franchise conditions. The question facing the profession is not whether AI will change library work, it already has. The question is whether librarians will organise collectively to demand systems serving public good rather than private extraction, build alternatives to vendor monopolies, and refuse to operate franchises that betray professional values.
The undergraduate student who discovered fabricated references in content supplied by vendors asked: ’Can I trust the library’s resources?’ How the profession answers, through collective action, not individual accommodation, will determine whether academic libraries remain vital knowledge institutions or become technical interfaces for corporate platforms.
The author thanks the two anonymous reviewers of Information Research, whose careful and substantive engagement across two rounds of review materially strengthened the paper’s framework, case study development, and international scope. Thanks are also owed to Professor Thomas Wilson and the editorial team at Information Research for their stewardship of the review process, and to copy-editor Joanna Adewunmi for her assistance during production. This work also benefited from discussion at the 2025 Minitex Technical Services Symposium.