In response to growing concerns and ongoing discussions about the rapid integration of AI in library services across the library field, Library Futures’s AI Policy Working Group is developing a model AI (artificial intelligence) policy and guidelines to inform collection development, reference, discovery, and services in public libraries. This project is part of Library Futures’s 2025-26 Research Network. Our Working Group conducted a survey of public library workers to identify and better understand their specific concerns and areas of need. For this project, we chose to focus on public libraries. Although public libraries vary greatly in size, location, and community context, focusing on public libraries as a distinct library setting allows us to tailor our guidelines to the conditions and practices that many public libraries share.

Our group recognizes there is significant institutional variation even among public libraries. Therefore, gathering direct feedback from public library workers helps ensure our guidelines reflect practitioners’ expressed concerns and needs, specific to their library settings, while making the project’s outcome relevant and adaptable across different public libraries.

In this post, we share key findings from our Spring 2026 survey of public library workers about their experiences with and perspectives on the need for and usefulness of AI policies in public libraries. The findings show a low rate of AI policy adoption despite high demand for one, specific areas of policy needs, and underlying tensions related to AI adoption in public libraries. The findings are based on responses from a total of 327 participants. We removed survey responses that did not meet Qualtrics quality filters (e.g., completeness, cache checks, and bot detection) for quality purposes. Although this sample cannot capture the perspectives of all public library workers, it provides a starting point to hear directly from them about their concerns, areas of need, and preferences for AI policy guidelines and resources.

Key Findings

1. Prevalence of an existing AI policy

Nine percent of respondents said their libraries had an AI policy. Slightly over half (53%) of respondents worked at public libraries that did not have an AI policy. About one-fifth (21%) indicated their libraries were in the process of creating such a policy. Seventeen percent were unsure whether an AI policy existed at their library. Image 1: Prevalence of existing AI policies bar graph. Note: “More than one AI policy” here refers to libraries having separate AI policies for different areas of library operations or services, e.g., collection development, human resources, internal staff use, and public services.

2. Usefulness of existing AI policies

To assess the usefulness of existing AI policies, we focused on the 29 survey respondents who indicated their library had at least one AI policy (28 with a single AI policy and 1 with more than one policy). We asked how useful existing AI policies are for different purposes. Using a scale of 1 (not at all useful) to 5 (extremely useful), respondents considered the usefulness of their library’s AI policy in relation to the following:

• Evaluating AI tools
• Understanding permitted uses of AI by staff
• Educating patrons about AI
• Providing staff with training and/or professional development
• Considering ethical and privacy concerns

(see Image 2 and Image 3 below)

The majority of respondents found their current AI policy to be “not at all useful” for evaluating AI tools, educating patrons about AI, providing staff with training and/or professional development, and considering ethical and privacy concerns. Understanding permitted uses of AI by staff was the only category for which the majority of respondents found their current AI policy to be useful, while even that category had the low ranking of “slightly useful.”

Image 2: Usefulness of current AI policies by count per use category

Image 3: Usefulness of current AI policies by average score per use category. Note: Responses were rated on a five-point scale, where 1= Not at all useful and 5= Extremely useful.

The category of greatest reported usefulness was “understanding permitted uses by staff,” followed by “considering ethical and privacy concerns.” The categories of least usefulness were: “educating patrons about AI” and “providing staff training and/or professional development.” Ratings for “evaluating AI tools” were overall low, though two respondents rated this category as most useful (rating of 5).

3. Gaps in existing AI policies

Among the 29 respondents whose libraries had an AI policy, 15 replied to the question “What gaps in the AI policy have you noticed? Select all that apply.” The selected categories, listed from most to least common, were:

• patron usage (12)
• evaluation or approval process for new AI tools (10)
• ethical considerations (9)
• privacy protections (9)
• staff training (8)
• permitted uses by staff (7)
• other (5)
• planned regular evaluation and update of policy (4)

4. Need for AI policies

When we asked staff at institutions without an AI policy about their perceived need for an AI policy, the overwhelming majority believed their library needed one. Those who viewed an AI policy as unnecessary usually indicated they did not see library workers and/or patrons using these technologies, did not view AI as appropriate for their library work, and/or were in a small or rural community where AI did not appear to be of great interest to the community members they served.

Respondents who saw a need for an AI policy believed it would provide clear guidelines that benefited library workers and library patrons. In open-text responses, 256 respondents noted varied approaches to AI within their organizations. These varied approaches to AI within workplaces were influenced partly by different understandings of AI and of ethical AI use. Reasons for needing an AI policy included:

• Protecting patron privacy
• Improving information integrity
• Guiding collection development and vendor relations
• Labeling and removing AI-generated content in subscription services and products
• Increasing AI literacy among both library workers and patrons
• Negotiating library contracts with vendors
• Understanding the impact of AI on library job security

Although most respondents saw a need for an AI policy, their open-text responses revealed contrasting views on the appropriate place for AI in library work and services. These differences suggested that crafting an AI policy would be challenging in many libraries. Several participants commented on this difficulty and on their uncertainty about how their library would engage in AI policy development. Many respondents described AI as antithetical to the purpose and values of public libraries and believed AI use should be avoided. Some viewed AI as a valuable tool for library work and/or for patrons. Others did not take a clear stance on whether AI should have a role in libraries but believed its prominence requires clear library policies.

Another commonly expressed concern was that library policies should align with municipal and state governance policies. Some stated that policy development and adoption might need to be driven by their municipalities’ work. Another noted issue was a perceived divide between, on one hand, upper library administrators and managers promoting AI use and, on the other, library staff concerned about AI adoption and use in library work.

5. Need for AI policy resources

When asked to rate what types of resources would be most useful for their libraries in developing AI policies, respondents overall indicated that the most useful information, listed in order of priority, would be:

• best practice recommendations for the general use of AI in public libraries
• best practice recommendations for specific AI uses, and
• several policy templates designed for specific AI topics (e.g., evaluating tools, staff training, privacy protections).

6. Points of tension

As noted above, respondents’ perspectives on the ideal role of AI adoption and usage in public libraries varied. Areas of tension included:

• Concerns about information integrity and trust in libraries: Many respondents expressed concerns about both the quality of AI-generated content and the consequences of libraries’ engagement with AI. One issue was AI slop sold by library vendors and potentially included in library collections, and the difficulty of identifying and addressing low-quality AI-generated materials. Relatedly, respondents were concerned about accidental engagement with AI, such as unknowingly purchasing a book with an AI-generated cover. Some worried that library use of AI could negatively affect public trust in libraries.
• Need for education, training, and institutional guidance: Many respondents remarked on the need for more education, staff training, and clearer institutional guidance on AI, including on different forms and uses of AI and implications. These individuals wanted clearer direction on permitted uses of AI in both patron services and internal operations. For example, respondents questioned whether AI could be used in reference services or to support language access, or whether it could be used internally for tasks such as note-taking. Concerns about AI-generated meeting notes were especially significant, since such notes could become part of the public record.
• Role of geography, library size, and digital infrastructure: Respondents, especially those in small and/or rural libraries, called for greater attention to the implications of AI adoption. They wanted guidance that reflects the realities of rural/small library staff wearing multiple hats. The call for attention to local context shows how infrastructure, staffing, funding, community needs, and community makeup and size may influence the implementation of AI across different library settings.
• Questions about the value of AI adoption: Underlying many of these responses was a broader skepticism and uncertainty about both the inevitability and the value of AI adoption. Some respondents firmly rejected the use of AI. Others occupied a grey area, expressing concern that adoption might be inevitable, even when its benefits remained uncertain. Regardless of where respondents fell on this spectrum, many who provided open-ended responses worried about AI’s environmental impact, privacy, surveillance, financial costs, and output quality.

Conclusion

Taking these findings into account, the Library Futures AI policy working group will produce:

• General AI policy guidelines that provide practical tips; pros, cons, and implications of AI adoption or non-adoption; and considerations when developing an AI policy
• A model AI policy or policies that can be adapted for individual libraries’ local contexts
• Recommendations for negotiations with vendors that prioritize community needs and library values like privacy and information integrity, and that require vendors to provide opt-outs for libraries and library users.
• Recommendations for collection development and collection development policies on topics like crafting or updating collection development policies, identifying AI-generated materials and taking appropriate action, appropriate labeling and metadata for any AI-generated content, and related staff training
• A bibliography of relevant resources

Though the study’s scope is limited, there seems to be a clear consensus that public libraries want more guidance and resources for navigating the AI landscape. Library Futures’s forthcoming work will be one more tool to help libraries weigh important considerations and respond to their communities’ needs.