Editor’s Note: Today’s post is by Mike Olson, Assistant Professor and Cataloging Librarian at Murphy Library, University of Wisconsin-La Crosse. Building on “Classification as Colonization,” this examination reveals how vendor consolidation and AI automation are systematically eliminating the invisible labor that sustains library systems.

The recent layoffs at OCLC, the Dublin, Ohio-based nonprofit that manages the Dewey Decimal System and WorldCat, offer a stark glimpse into the precarious future of library and information labor. In July 2025, OCLC confirmed it reduced its central Ohio workforce by about 80 positions, citing “shifts in technical skill requirements, growing influence of artificial intelligence, and ongoing changes in higher education and libraries.”

The timing reveals a fundamental contradiction in how cataloging expertise is valued. While OCLC cites AI as justification for workforce cuts, the organization has also announced it is developing AI tools for metadata processing. Here lies the paradox of our moment: the same technological advances celebrated for their efficiency are erasing the human expertise that creates the high-quality metadata these systems depend on to function.

This contradiction becomes intelligible within the broader transformation of universities from public good institutions to market-driven entities focused on knowledge commodification. Library labor exemplifies what Maurizio Lazzarato identifies as “immaterial labor” — work that produces both informational content and cultural standardization, becoming directly productive for digital capitalism through vendor-controlled systems that aggregate and monetize standardized metadata.

The Invisible Infrastructure Under Attack

The devaluation of cataloging expertise becomes visible when institutions eliminate these positions during budget cuts. Northwestern University’s announcement of layoffs affecting 425 positions included Violet Fox, one of the field’s leading voices in critical cataloging and creator of the Cataloging Lab. Fox’s dismissal is particularly telling — her work on reparative description practices exemplifies exactly the kind of specialized cataloging knowledge that institutions are eliminating precisely when it’s most needed. Catalogers like Fox perform the invisible work of challenging colonial classification systems, ensuring that subject headings reflect community needs rather than vendor-supplied generic terms, making collections discoverable and accessible, and maintaining the metadata quality that enables scholarly research.

From Card Catalogs to Vendor Control

The nostalgia for card catalogs that occasionally surfaces in library discourse isn’t mere romanticism; it reflects a genuine loss of institutional autonomy. The physical card catalog represented direct institutional control over bibliographic organization, where local catalogers made decisions about subject headings, classification, and cross-references based on their collections and communities’ needs.

Today’s “online catalogs” increasingly function as interfaces to vendor-controlled discovery systems that aggregate metadata from multiple sources with minimal local oversight. Collection Discovery Indexes (CDIs) like EBSCO’s Discovery Service or ProQuest’s Primo Central dump millions of records from contracted vendors directly into library search interfaces, creating what appears to be a comprehensive catalog but actually represents a fundamental transfer of intellectual control from local institutions to corporate entities.

This transfer represents platform capitalism’s colonization of library infrastructure. Where card catalogs embodied local cataloging decisions, contemporary discovery systems operate through what Shoshana Zuboff identifies as the extraction architectures that gather user behavioral data while providing search results optimized for vendor algorithms rather than local collection strengths or community needs.

The metadata quality crisis in vendor-supplied records typifies this loss of control. Local catalogers historically ensured consistency and accuracy through direct engagement with materials. Vendor-supplied metadata often contains errors and generic subject headings that reflect automated processing rather than intellectual analysis. Yet libraries have a limited ability to correct these records without disrupting system updates that overwrite local modifications.

Vendor Consolidation and Legal Battles

The OCLC layoffs occur within the broader vendor consolidation that has fundamentally altered power dynamics within library ecosystems. Multiple legal battles illuminate these tensions. OCLC’s recent settlement with Clarivate over the MetaDoor dispute forced Clarivate to cease development and permanently delete work product that incorporated WorldCat records. As The Scholarly Kitchen reported, this lawsuit revealed deep tensions over who controls bibliographic metadata created through collaborative library labor. Even more revealing is OCLC’s ongoing litigation against Anna’s Archive, accused of scraping WorldCat’s data.

What’s particularly striking is how OCLC’s litigation strategy positions libraries. While OCLC doesn’t typically sue libraries directly, its lawsuits create a chilling effect that constrains how libraries can use their own contributed metadata. Libraries find themselves contractually restricted from sharing WorldCat records outside OCLC-approved channels, sometimes unable to freely reuse catalog data they themselves created and contributed to the collective database.

This reveals a deeper tension: WorldCat emerges from decades of collective cataloging labor by thousands of librarians, yet OCLC asserts proprietary control over this collaboratively created resource. Libraries become trapped in a system where their intellectual labor is incorporated into a collective good, but access to that resource remains subject to corporate control and contractual restrictions.

AI’s False Promise: From Cataloging to Discovery

The promise of AI in library cataloging is seductive: faster processing, reduced manual entry, freeing workers for “more important” tasks. Yet the Library of Congress’s recent experiments found that Large Language Models scored only a 26% F1 when predicting Library of Congress Subject Headings, while subject classification models achieved only 35% accuracy. These results reveal a fundamental shift in cataloging labor from record creation to training, evaluating, and curating algorithmic outputs—catalogers now find themselves providing feedback and quality review for systems that cannot replicate the human expertise, cultural knowledge, and contextual understanding that quality metadata requires.

The disconnect between AI promises and library reality was already starkly apparent at the 2024 Ex Libris Users of North America (ELUNA) conference. While Ex Libris staff delivered polished presentations touting AI integration into every facet of their products, from customer service to cataloging, user sessions revealed a different story entirely. Librarians shared workarounds for basic functionality that had remained broken for years, struggling with vague documentation and systems that failed to deliver even simple results like getting resources to appear prominently in search results. The tension was palpable during Q&A sessions, where librarians voiced disinterest in AI features while grappling with fundamental system problems.

More troubling is how AI discovery interfaces bypass cataloging work entirely. Aaron Tay’s investigation exposed how Clarivate’s Summon Research Assistant and Primo Research Assistant refuse to search terms like “Tulsa race riot,” “Gaza War,” or anything containing “massacre,” returning zero results from billion-record databases. Content-filtering layers designed for social media chatbots now “quietly block controversial topics from being searched” in academic library systems, undermining the careful subject analysis that catalogers perform to ensure comprehensive access to scholarly materials.

This represents two simultaneous threats to library infrastructure: AI cataloging systems that require extensive human oversight while promising to eliminate catalogers, and AI discovery interfaces that replace structured metadata searching with chatbot interactions that can silently censor topics. Where traditional library catalogs surface resources through cataloger-assigned subject headings and classification numbers, AI discovery tools depend on algorithmic interpretation that can simply refuse to search—making catalogers’ intellectual work invisible while constraining access to the very materials they’ve carefully described and organized.

Implications for Scholarly Communication

The displacement of cataloging expertise carries profound consequences for how scholarly knowledge circulates. Publishers increasingly find themselves dependent on vendor algorithms rather than subject-specialist catalogers to connect their content with researchers. Where human catalogers historically ensured that resources could be discovered through multiple access points—alternative subject headings, cross-references, and classification schemes that reflect disciplinary nuances — AI discovery systems operate through black-box algorithmic interpretation that prioritizes vendor metrics over scholarly relevance.

This shift fundamentally alters the relationship between content creators and researchers. Vendor-supplied metadata often contains errors and generic subject headings that reduce precision in academic search results, making it harder for scholars to locate relevant sources and for publishers to demonstrate usage and impact. The transformation from locally-controlled catalogs built on human expertise to vendor-mediated platforms means that discoverability increasingly depends on corporate decisions about interface design, search algorithms, and content filtering rather than intellectual analysis of scholarly materials.

Most troubling is how AI implementation can actively constrain academic inquiry. When discovery systems can refuse to search entire categories of terms or topics, they undermine the foundational principle that researchers should have comprehensive access to relevant materials. As Jay Singley warns, “Uncritically adopting AI tools in discovery systems will perpetuate, if not exacerbate, existing biases and suppression of minoritized people”—a concern that extends beyond representation to fundamental questions about what knowledge remains discoverable in an algorithmic environment.

The Price of Automation

The crisis facing library labor today extends the colonial dynamics explored in “Classification as Colonization.” Just as classification systems can perpetuate epistemic violence through seemingly neutral organizational choices, labor policies can systematically devalue and eliminate the human expertise that makes libraries function.

The 80 workers laid off at OCLC, the elimination of Violet Fox’s position at Northwestern, and thousands of other displaced catalogers represent more than statistical casualties—they embody decades of accumulated expertise in organizing human knowledge that, once lost, cannot be easily recovered. When AI systems can silently block searches for “Gaza War” or “Tulsa race riot” while achieving only 26% accuracy on subject classification, we witness the replacement of human intellectual judgment with corporate algorithmic control that serves surveillance capitalism rather than scholarly inquiry.

Understanding this crisis requires theoretical frameworks that connect institutional transformation, technological change, and worker experience within broader patterns of academic capitalism and surveillance capitalism’s penetration of knowledge work. As libraries become more central to digital knowledge production and commodification, cataloging labor becomes simultaneously more visible, more technologically mediated, and more affected by commercial and algorithmic logics.

As Langdon Winner’s foundational question — “Do Artifacts Have Politics?” — reminds us, library technologies embody social relations and power structures rather than neutral tools. The challenge isn’t whether AI will change library work, it’s whether those changes will serve the public good or private profit, whether they’ll enhance human expertise or eliminate it, whether they’ll democratize knowledge or concentrate control in vendor platforms that can censor at will.

The stakes could not be higher. In an era of information abundance and epistemic chaos, libraries serve as crucial sites for democratic knowledge practices. The catalogers and information workers being eliminated today are the same workers who challenge colonial classification systems, create community-responsive metadata, and ensure that controversial but essential topics remain discoverable. Protecting these workers isn’t just about labor justice, it’s about preserving the infrastructure of informed citizenship itself. The question isn’t whether libraries will survive digital capitalism’s transformation, but whether they can become sites of resistance to it.