The AI Revolution: How Chatbots Are Transforming the Library Reference Interview

Gretchen GardellaNo CommentsBlog, Liblime

The reference interview has long been the cornerstone of library service, representing a carefully choreographed dialogue between the librarian and the patron designed to uncover patrons’ true information needs. According to the Online Dictionary for Library and Information Science, a reference interview is “the interpersonal communication that occurs between a reference librarian and a library user to determine the person’s specific information need(s), which may turn out to be different from the reference question as initially posed” (1). This structured conversation, refined over decades of professional practice, allows librarians to move beyond surface-level queries to identify underlying research goals. Artificial intelligence chatbots such as ChatGPT and Claude, along with library-specific tools, are fundamentally reshaping traditional interactions, presenting both unprecedented opportunities and significant challenges for the future of reference services.

The reference interview emerged as a formalized practice in the mid-twentieth century, codifying what skilled librarians had always known: patrons rarely articulate their complete information needs in initial queries. The traditional process encompasses five main areas: approachability, interest, listening and inquiring, searching, and follow-up (2). A student asking for “books about dogs” might actually need sources on animal-assisted therapy for a psychology paper, while someone requesting “climate change data” could be writing anything from a high school essay to a doctoral dissertation. These structured steps are designed to put users at ease and ensure they have clearly explained their requirements, with the often-overlooked final step being verification that the information provided was indeed what the user needed (1).

AI chatbots are now performing many functions traditionally handled through reference interviews, and implementation statistics demonstrate their growing influence. The University of Calgary Libraries, which implemented a multilingual AI reference chatbot in 2021, reported that the chatbot deflected 50 percent of questions from human-managed live chat, thereby freeing 1.5 full-time equivalents of reference staff to support higher-level tasks (3). At San Jose State University Library, a chatbot deployed in spring 2021 initially logged approximately 44 monthly interactions, increasing to approximately 137 by spring 2022 (4).

The appeal is understandable. AI chatbots are available twenty-four hours a day, never exhibit frustration with repetitive questions, and can simultaneously assist unlimited users. They eliminate the social anxiety that some patrons experience when approaching reference desks, which is particularly important for neurodivergent individuals or those from cultures where

Screenshot 2026 02 14 140605

https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2024/01/ai-companies-uphold-your-privacy-confidentiality-commitments

questioning authority figures feels uncomfortable. These systems can process questions instantly, search vast databases, and provide tailored responses without wait times or scheduling constraints.

However, AI chatbots currently struggle with aspects of reference work that librarians consider essential. A study by Yang (2024) that tested ChatGPT with twenty-two reference questions found that while ChatGPT excelled in information retrieval in some areas, it was not comparable to a reference librarian in others (5). The research revealed a significant problem: ChatGPT recommended numerous imposter articles and fabricated fake citations for them. When questioned about this behavior, ChatGPT stated that, as an AI model, it aims to provide helpful information but can sometimes make mistakes due to limitations in its training (6).

Another comprehensive study by Lai (2023) analyzed ChatGPT’s performance using different question types and difficulty levels, using the Reference Effort Assessment Data Scale. The overall assessment found that ChatGPT’s performance was fair, but it performed poorly in information accuracy. ChatGPT scored highest on questions about facilities and equipment-related questions but lowest on electronic resource access problems. The study concluded that ChatGPT was weak at answering advanced research questions, complex inquiries, and known-item searches related to a specific local environment (7).

The problem extends beyond accuracy to encompass issues of information literacy and the development of critical thinking. Reference librarians don’t simply answer

Screenshot 2026 02 14 140801

https://journals.sagepub.com/doi/10.1177/09610006231224440

questions; they teach patrons how to evaluate sources, understand database structures, and develop research strategies that transfer across projects. When librarians guide students through database selection, explain why certain keywords yield better results, or demonstrate how to trace citations backward through the literature, they’re building research competence that serves patrons beyond any single interaction. Li and Coates (2024) found that while ChatGPT could effectively provide general guidance on locating information resources, optimized AI online chatbots “still have a long way to go to meet the dynamic needs of faculty and students in the ever-changing academic learning environments” (8).

Privacy concerns add another dimension to this transformation. Traditional reference interviews, although not legally privileged like attorney-client communications, are governed by professional ethics codes that emphasize client confidentiality. The American Library Association’s Code of Ethics explicitly protects patron privacy in library interactions, but AI chatbot conversations often involve data collection and storage practices that complicate these protections. The Federal Trade Commission has warned that AI companies have “a continuous appetite for data to develop new or customer-specific models or refine existing ones,” noting that customers may reveal sensitive or confidential information when using AI models, creating risks that such data could be inappropriately accessed or leaked (9).

The integration of AI into reference services also raises questions about algorithmic bias and equitable access to information. Reference librarians receive training in serving diverse communities and can recognize when their own biases might influence service quality. AI systems, however, reflect biases present in their training data and design choices made by developers who may not understand library values. A qualitative study examining stakeholder perspectives on chatbot adoption in academic libraries found that while the majority favored chatbot integration, “perceived risk” with respect to employing chatbots was high, particularly regarding the chatbot’s ability to develop appropriate responses and understand user intent (10).

Despite these challenges, forward-thinking libraries are developing hybrid models that leverage AI capabilities while preserving human expertise where it matters most. These approaches position chatbots as first-line responders for straightforward questions while seamlessly escalating complex queries to human librarians. Aboelmaged et al. (2024) conducted an integrative literature review examining the evolving role of conversational AI chatbots in library services, identifying key themes including technological evolution, user experience, and challenges faced (11). Their research suggests that the most effective implementations use hybrid models where AI handles repetitive tasks, allowing librarians to focus on in-depth research support.

Some libraries are also using AI tools to enhance rather than replace reference interviews. Chatbots can pre-screen patrons by gathering basic information about research topics, assignment requirements, and prior search attempts, before connecting users with librarians who then have a richer context for conducting effective interviews. This model preserves the essential elements of human reference work while reducing time spent on information gathering, which machines can efficiently handle. Rodriguez and Mune (2022) documented San Jose State University Library’s experience deploying a chatbot using natural language processing and AI training for basic circulation and reference questions, noting that continuous maintenance of chatbot content and user testing were essential for facilitating user engagement (4).

The future likely involves reference librarians developing new competencies centered on AI literacy and hybrid service models. Rather than conducting every reference interview from scratch, librarians may increasingly review and refine chatbot-generated responses, teach patrons how to query AI systems effectively, and focus their expertise on complex research problems that require nuanced judgment. A 2025 study on augmenting knowledge access in next-generation digital libraries noted that institutions adopting AI assistants have observed notable improvements in user satisfaction, particularly when transparency is maintained—making it clear to users when they are interacting with AI rather than a human (12).

Screenshot 2026 02 14 142311

https://www.lisedunetwork.com/artificial-intelligence-ai-chatbots-for-library-reference-services/

Educational institutions are beginning to reshape library science curricula to prepare future librarians for this reality. Programs now incorporate training in understanding AI capabilities and limitations, evaluating algorithmic outputs, and the ethical implications of automated information systems alongside traditional reference interview techniques. The goal is to develop professionals who understand both the power and limitations of AI tools and can make informed decisions about when human judgment remains essential.

Critical questions remain about how this transformation will affect library values and mission. Will the efficiency of AI chatbots create pressure to reduce professional librarian positions, ultimately diminishing service quality when algorithms prove inadequate for complex needs? How can libraries ensure that AI tools serve all patrons equitably rather than amplifying existing inequalities? What happens to serendipitous discovery and the informal information literacy instruction that occurs during traditional reference interactions?

Chen (2023) argued that the ability and capacity of ChatGPT to answer questions not limited to physical information sources or specific library locations is at a suitable and

Screenshot 2026 02 14 144356

https://crl.acrl.org/index.php/crl/article/view/26102/34024

convincing level, but emphasized that “the need for human librarians will never disappear” because AI provides very general answers and cannot address queries specific to local library environments (13). This observation captures the essential tension: AI can handle routine inquiries efficiently, but the pedagogical relationship and contextual expertise that characterize excellent reference service remain fundamentally human endeavors.

The reference interview is not merely a transaction but a pedagogical relationship that empowers people to become more sophisticated information seekers. As AI chatbots increasingly mediate these interactions, librarians must actively shape implementation to preserve what makes reference services genuinely valuable while embracing technologies that can extend their reach and impact. The challenge lies not in resisting change but in ensuring that technological transformation serves library missions of equitable access, information literacy, and intellectual freedom rather than undermining them.

The reference interview is evolving, not disappearing, but its future form will require intentional design that centers human needs and library values. AI chatbots will undoubtedly handle an increasing share of reference work. Still, the question remains whether this shift will enhance or diminish the educational mission at the heart of library service. The answer depends on how thoughtfully libraries integrate these tools while maintaining their commitment to developing informed, critical thinkers capable of navigating an increasingly complex information landscape.

 

References

  1. Reference interview. (2025, November 2). In Wikipedia. https://en.wikipedia.org/wiki/Reference_interview
  2. National Archives. (n.d.). Guidelines for a successful reference interview. https://www.archives.gov/files/boston/volunteers/reference-interviews.pdf
  3. OCLC Research. (2024, November 20). Lessons learned from implementing an AI reference chatbot at the University of Calgary Library [Webinar]. https://www.oclc.org/research/events/2024/implementing-an-ai-reference-chatbot.html
  4. Rodriguez, S., & Mune, C. (2022). Uncoding library chatbots: Deploying a new virtual reference tool at the San Jose State University Library. Reference Services Review, 50(3–4), 392–405. https://doi.org/10.1108/RSR-05-2022-0020
  5. Yang, S. Q. (2024). ChatGPT: Unleashing the power of conversational AI for library reference services. International Journal of Librarianship, 9(1), 109–115. https://doi.org/10.23974/ijol.2024.vol9.1.375
  6. Cox, K. (2025, February 13). What can ChatGPT do in the library? Katina Magazine. https://katinamagazine.org/content/article/resource-reviews/2025/what-can-chatgpt-do-in-the-library
  7. Lai, K. (2023). How well does ChatGPT handle reference inquiries? An analysis based on question types and question complexities. College & Research Libraries, 84(6). https://crl.acrl.org/index.php/crl/article/view/26102/34024 
  8. Li, L., & Coates, K. G. (2024). Academic library online chat services under the impact of artificial intelligence. Information Discovery and Delivery, 53(2), 192–205. https://www.emerald.com/idd/article-abstract/53/2/192/1241777/Academic-library-online-chat-services-under-the?redirectedFrom=fulltext
  9. Federal Trade Commission. (2024, January). AI companies: Uphold your privacy and confidentiality commitments. https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2024/01/ai-companies-uphold-your-privacy-confidentiality-commitments
  10. Prakash, T., & Kausalya, S. (2025). The role of AI chatbots in academic libraries: Opportunities and challenges. Journal of Emerging Technologies and Innovative Research (JETIR), 12(10). https://www.jetir.org/papers/JETIR2510155.pdf 
  11. Aboelmaged, M., Bani-Melhem, S., Al-Hawari, M. A., & Ahmad, I. (2024). Conversational AI chatbots in library research: An integrative review and future research agenda. Journal of Librarianship and Information Science, 57(2). https://doi.org/10.1177/09610006231224440
  12. Hasan, N., et al. (2025). Augmenting knowledge access: The role of artificial intelligence in next-generation digital libraries. International Journal of Research in Library Science, 11(1). https://www.ijrls.in/wp-content/uploads/2025/09/ijrls-1942.pdf
  13. Saeidnia, H. (2023). Using ChatGPT as digital/smart reference robot: How may ChatGPT impact digital reference services? Information Matters. https://informationmatters.org/2023/05/using-chatgpt-as-digital-smart-reference-robot-how-may-chatgpt-impact-digital-reference-services/