Librarians Pushing Back Against the AI Boom

Gretchen GardellaNo CommentsUncategorized

Professional Skeptics

Librarians have a long record of adopting new information technology, from online catalogs to e-books and public internet access. Generative AI has been different. As chatbots and AI features spread through search engines, research databases, and campus software, many library workers have responded with organized skepticism rather than enthusiasm.

A survey of the publishing supply chain by the Book Industry Study Group (BISG), discussed at an April 2026 webinar, found that about a third of library respondents did not use AI and had no plans to, compared with 20% across the full dataset. Roughly 34% of librarians described themselves as ethically opposed to its use [1]. Those numbers do not describe a profession in unanimous revolt, and many librarians sit somewhere in between. But they point to resistance that stands out against the wider rush toward AI, and understanding it means looking at what librarians say the technology is doing to their daily work, their collections, and their professional values.

Chasing Books That Don’t Exist

The most visible frustration is practical. Chatbots can produce citations that look correct, with real author names, real journal titles, and plausible page numbers, for articles and books that were never written. In January 2026, Emilia Marcyk, head of Reference and Discovery Services at Michigan State University Libraries, told WKAR that librarians are fielding more requests to locate such sources, and that proving a reference does not exist is a uniquely hard task [2]. She said fabricated references have grown harder to spot as the tools get better at sounding plausible, and that some patrons now arrive with an answer already in mind that they simply want confirmed [2].

Writing in the March/April 2026 issue of Computers in Libraries, Marydee Ojala argued that chatbots’ habit of inventing not just articles but entire journals complicates librarians’ work, wastes professional time, and that verifying citations is slow regardless of the method used [3]. She also noted that bad information long predates AI; in her view, generative tools amplify an old problem rather than create a new one [3]. Librarians in the BISG survey described related burdens, including catalogs filling with low-quality AI-generated books and staff hours spent identifying and removing them [1].

Vendors, Privacy And Bots At The Door

A second set of concerns involves the products libraries already pay for. Database and discovery vendors have been adding AI research assistants and summaries to their platforms. In a 2026 paper in Endnotes, the journal of the American Library Association’s (ALA) New Members Round Table, Cal Poly Pomona librarians Amber Wu and Michelle S. Navarro argued that libraries risk being cast in passive roles when vendors or campus leaders set the timeline and value of AI tools before any library-centered evaluation happens [4]. They added that vendor demonstrations often fail to explain how a product differs from freely available chatbots [4].

Privacy sharpens these worries for a profession whose ethics code commits it to patron confidentiality. ALA’s 2026 AI guidance urges libraries to check whether AI features are switched on by default, whether they can be disabled, and whether searches and prompts are logged or used to train models [5].

Libraries and archives also face AI’s appetite for data. In August 2026, Fast Company reported that crawlers linked to Meta and Alibaba nearly knocked offline the LGBT History Project, a volunteer-run UK archive with more than 50 million views that is preserved by the British Library, slowing the site and raising its hosting costs [6]. For institutions that spent years making collections freely available, such episodes can feel like the public’s resources being drained for private gain.

Refusal As A Professional Choice

These grievances are now shaping formal policy. In July 2026, the ALA Council adopted “Guidance on the Use of Artificial Intelligence in Libraries,” written by a 30-member working group and meant as a living document to help libraries draft local policies [7]. The guidance does not reject AI outright; it says AI may support instruction, accessibility, and operations when it serves a clear public-service purpose and preserves human judgment [5]. But it also states that a considered decision to delay, remove, or decline a product is an affirmative professional choice, not a failure of adoption [5]. It recommends that libraries not use AI to justify staff cuts or replace reference and readers’ advisory roles, weigh energy and water use in procurement, and recognize that declining to buy mainly AI-generated books is a legitimate collection decision [5].

A more explicit “critical refusal” current has also grown. At the LILAC information literacy conference in Sheffield in spring 2026, librarians Salma Abumeeiz and Jamie Hazlitt challenged the assumption that generative AI is inevitable and argued that refusing it is both possible and necessary, while acknowledging the real pressure librarians face to adopt it [8]. In January 2026, researchers at the University of North Carolina at Greensboro opened a national survey to document refusal, skepticism, and non-compliance among academic library workers [9]. Library scholar R. David Lankes, who has run focus groups with “never AI” librarians, has said their resistance tends to come from people who understand the technology well, not from unfamiliarity [1].

The Case For Engagement

Many librarians reject the idea of a fight altogether. Adoption is widespread in academic libraries: in the 2025 Ithaka S+R survey of U.S. library leaders, summarized by Elsevier in July 2026, only 24% said their library does not use AI for internal work or services, and leaders broadly expected demand for AI tools to grow [10]. Among the barriers cited, the most common were low priority relative to other strategies (50%) and limited staff AI skills (48%), which are practical rather than ethical limits [10].

Supporters point to concrete benefits. Lankes described a University of Texas music collection of thousands of uncataloged recordings that became discoverable after AI generated provisional records tagged with confidence scores, with the least certain flagged for human review [1]. Medical librarians at a 2026 Southeastern Conference AI library summit described “leaning in” as a way to keep professional agency, treating AI as a tool with real limits and using it to teach critical thinking [11].

The view is echoed internationally. A 2026 study of 43 librarians’ views in a developing country concluded that AI is more likely to serve as a collaborative tool than to replace librarians [12]. At a South African Librarians’ Day event in Johannesburg, a city librarian argued that those who master AI will be better equipped to serve their communities [13]. Researchers also appear to want help: in Elsevier’s 2026 “Researcher of the Future” report, only 45% said they had adequate AI training, and just 32% believed their institution governed AI well [14].

Common Ground

Taken together, the evidence suggests librarians are less united against AI than wary of how quickly, and on whose terms, it is arriving. Critics and enthusiasts tend to agree on several points: AI output should be verified, patrons deserve to know when AI is in use, and human expertise should not be quietly automated away. ALA’s guidance reflects that middle ground, asking libraries to teach AI literacy without shaming people who use, avoid, or distrust the tools [5]. Whether the profession’s resistance slows AI adoption or steers it toward something more accountable, librarians have made themselves one of the few groups asking, tool by tool, whether a given technology actually serves the public.

Sources

  1. Friedman, J. (2026, April 20). AI and libraries: Why librarians may become arbiters of reality. Jane Friedman. https://janefriedman.com/ai-and-libraries-why-theyre-like-canaries-in-the-coal-mine/
  2. Pohl, S. (2026, January 6). Michigan librarians say they’re being asked to find non-existent references cited by AI chatbots. WKAR Public Media (via WCMU). https://www.wcmu.org/local-regional-news/2026-01-06/michigan-librarians-say-theyre-being-asked-to-find-non-existent-references-cited-by-ai-chatbots
  3. Ojala, M. (2026, March/April). Librarians, we have a hallucination problem. Computers in Libraries. https://infotoday.com/cilmag/mar26/Ojala–Librarians-We-Have-a-Hallucination-Problem.shtml
  4. Wu, A., & Navarro, M. S. (2026). Reflections on the adoption and integration of AI tools in academic libraries. Endnotes: The Journal of the New Members Round Table, 14(1), 63–71. https://journals.ala.org/index.php/endnotes/article/download/8805/12087
  5. American Library Association, Artificial Intelligence Policy Working Group. (2026). Guidance on the use of artificial intelligence in libraries (ALA CD#44.2). https://www.ala.org/sites/default/files/2026-06/ALA%20CD%2044.2%20AI%20Guidance%20Document%20-%20Final.pdf
  6. Stokel-Walker, C. (2026, August). AI crawlers from Meta and Alibaba almost took down a volunteer-run LGBT history archive. Fast Company (Fast Company Brasil edition). https://fastcompanybrasil.com/tech/bots-de-ia-da-meta-e-alibaba-quase-tiraram-do-ar-um-arquivo-historico-lgbtqia/
  7. American Library Association. (2026, July 16). ALA Council adopts Guidance on the Use of Artificial Intelligence in Libraries [Press release]. https://www.ala.org/news/2026/07/ala-council-adopts-guidance-use-artificial-intelligence-libraries
  8. Abumeeiz, S., & Hazlitt, J. (2026, March 30–April 1). Opting out of inevitability: Reflections on critical refusal of generative AI in academic libraries [Conference session]. LILAC 2026, University of Sheffield. https://www.lilacconference.com/events/2026/opting-out-of-inevitability-reflections-on-critical-refusal-of-generative-ai-in-academic-libraries
  9. University of North Carolina at Greensboro, Information, Library and Research Sciences Department. (2026, January 15). AI refusal in academic libraries – survey now open. Society of American Archivists Connect. https://connect.archivists.org/discussion/ai-refusal-in-academic-libraries-survey-now-open
  10. Elsevier. (2026, July 6). Four key findings in the 2025 Ithaka S+R librarian survey: An industry under pressure. Elsevier Connect.https://www.elsevier.com/connect/key-highlights-from-the-2025-ithaka-s-r-us-library-survey-a-sector-under-pressure
  11. Camarillo, L. A., & Martin-Klumpp, A. (2026, May 6). Lean on me: Supporting critical thinking using databases and AI [Video recording]. 2026 Southeastern Conference (SEC) AI Library Summit, Mississippi State University. https://scholarsjunction.msstate.edu/sec-ai-2026/13
  12. Chigwada, J., & Ngulube, P. (2026). Will artificial intelligence replace human librarians? Views from librarians in a developing country. South African Journal of Libraries and Information Science. https://sajlis.journals.ac.za/pub/article/view/2505
  13. City of Johannesburg. (2026, July). Noordgesig Library marks South African Librarians’ Day with focus on ethical AI. https://joburg.org.za/media_/Newsroom/Pages/2026-News-Articles/Noordgesig-Library-marks-South-African-Librarians-Day-with-focus-on-ethical-AI.aspx
  14. Elsevier. (2026, April 29). Five ways libraries can support researchers in the AI era. Elsevier Connect. https://www.elsevier.com/connect/five-ways-libraries-can-support-researchers-in-the-ai-era