Artificial intelligence (AI) is one of the most frequently discussed technology trends in libraries today. AI impacts multiple areas of the library including resources, collections and user experience. AI tools are entering the market at a rapid pace, with many publishers introducing their own built-in solutions, some of which come at a steep cost. These tools are designed to enhance the research experience by offering features such as automated summarization and improved discovery of relevant materials. This technology also promises to transform library management by streamlining workflows and improving access to resources.

Major resource vendors, like Elsevier and EBSCO, have developed AI resources aimed at helping users with resource utility. Summarization features and suggested related article linking are one of the most prevalent types of AI tools in the library space. More recently, upcoming AI library technology focuses on metadata enrichment in the form of chatbots that can help users with resources or library pages.

Despite the potential benefits with AI tools, a number of ongoing concerns remain. One area of concern is that search results are often irreproducible (Patterson, 2025). AI databases may lack transparency in how search strings are processed and expanded. This can be challenging for users when trying to recreate a search strategy or replicate a result list. Additional issues include link resolver compatibility and the complexity of managing access to resources that require individual user logins. AI resource enhancements or standalone platforms may come at an additional cost. Also of concern is that AI enhancements that are included in database subscriptions may not provide library personnel administrative controls, requiring users to encounter or interact with the tool by default.

Research at East Carolina University Libraries on AI Tools in Library Collections

In early spring 2025, the Laupus Health Sciences Library decided to distribute a brief survey to evaluate the current landscape of AI in health sciences library collections. The survey was distributed to the Medical Library Association Collection Development Caucus and was designed to highlight librarians� attitudes, concerns, and purchasing behaviors of independent AI tools such as ClinicalKey AI, Scopus AI, or Dyna AI.

Respondents identified several current concerns with available AI products, including cost, copyright issues, gaps in information literacy skills, hallucinations, and the perception that some tools still feel largely �beta-like� in their development. Relevancy and trustworthiness were also frequently cited issues, which may be partly attributed to certain products limiting searches to selected publisher-provided journals. Additional concerns included the lack of reliable linking to full-text resources, which can prevent users from accessing complete texts, as well as difficulties with how these systems interpret search queries. The majority of survey respondents indicated that they did not currently provide access to AI tools. However, those that did offer access to AI tools, reported providing the following to their patrons: OpenEvidence, TDNet AI, VisualDX�s DermExpert, and Scite. When asked if their library intended to purchase AI tools in the future, respondents most frequently indicated that they were unsure if they would purchase these types of tools.

More recently, East Carolina University (ECU) Libraries designed and distributed a multi-institutional survey that examined how libraries are approaching the evaluation, licensing, adoption, and policy development of AI tools. The study took a more in-depth approach to determine how libraries evaluate and manage AI technologies, assessing their policies and identifying management roles of these types of resources. Findings from this study indicated that the majority had no formal evaluation criteria, but the majority were currently developing an evaluation process. Decision-making about proceeding with and purchasing AI tools and products varied; respondents identified library directors, technology departments, and library/technology committees as the decision makers. Responses also indicated that the criteria used most frequently when considering the implementation of an AI tool was cost, accuracy and reliability, data privacy, and alignment with library values.

Evaluating AI Tools at ECU Libraries

With the proliferation of available tools and added AI features, East Carolina University Libraries set out to develop an evaluation method that could effectively identify the strengths, limitations, and practical implications of each option. To navigate the growing number of available products, we determined that a comparison matrix would provide a useful framework for evaluating and comparing these resources.

ECU Libraries developed our matrix based on an approach from Loyola Marymount University�s William H. Hannon Library, known as the REACT Framework (William H. Hannon Library, 2024). Criteria examined with the REACT Framework matrix include relevancy, ease of use, DEIA components, currency, transparency and accuracy. These are scored using a 1-4 scale. The libraries decided to utilize the REACT Framework for our current evaluation tool and added other key information points. We also included information on the ability to opt in or opt out of the tool, content sources, cost, security and privacy, data collection practices, integration compatibility, and support, as well as features and claims that the tool makes. While we are still determining the most effective ways to apply and refine this framework, it has already proven helpful in facilitating more systematic comparisons among AI tools and identifying key considerations during the evaluation process. As libraries continue to explore the growing number of AI technologies and applications, the need for structured evaluation methods such as a criteria evaluation matrix becomes increasingly important.

Conclusion

Artificial intelligence in the library field is evolving at a rapid pace. Beyond tools provided from library vendors, AI offers the potential to reduce workload by offering the opportunities for �creating advanced automation systems, personalized information delivery, real-time catalog updates, automatic indexing, and metadata generation� (Jhan, 2025). Whether implemented through vendor-provided platforms or locally developed solutions, libraries have the opportunity to unlock the full potential of AI across a wide range of applications and technologies. As AI continues to evolve, libraries will play an important role in evaluating and integrating these tools in ways that support their institutional goals and user needs.

References

Jhan, P., Sreekumar, M., & Kuriakose, R. (2025). Enhancing library services with artificial intelligence: A framework for an automated news delivery system. IFLA Journal, 51(3), 836-848.

Patterson, B. (2025). Can AI help with that? the limitations of AI tools for information discovery, search and reviews. Journal of Electronic Resources in Medical Libraries, 22(1-2), 56�59. 10.1080/15424065.2025.2496622

William H. Hannon Library. (2024). Using the REACT Framework to Evaluate and Assess Generative AI Tools. Loyola Marymount University, William H. Hannon Library. libguides.lmu.edu/GAIL24/REACTFramework

Resources for Evaluating AI Tools

• Using the REACT Framework to Evaluate and Assess Generative AI Tools
• AI Tools for Research
• AI Competencies for Academic Library Workers
• Generative AI: Supporting AI Literacy, Research, and Publishing: Evaluating AI Tools & Outputs

Editor�s note: This article is based on a presentation given by the author at the 2025 Medical Library Association conference entitled �AI and Collections in Health Sciences Libraries: Bridging the Gap Between Publisher Tools and the Libraries.�

DCT Featured Article � June 9, 2026