The most useful research for understanding where any profession stands on AI isn’t always the most cited. Lo’s (2024) survey of 760 academic library employees in the United States, published in College & Research Libraries, names something the discourse around AI in education keeps obscuring: there’s a wide gap between recognizing AI’s value and being ready to use it. The same gap exists in K-12 and higher ed teaching, and the survey’s structure for measuring it transfers cleanly.
Half of Lo’s respondents see generative AI as beneficial to library services. About the same proportion think their library should invest in AI. But 70% report being unprepared to adopt these tools in the next twelve months. Lo summarizes the pattern: “the readiness gap in AI adoption uncovered by the study suggests a disconnect between understanding the potential of AI and the ability to harness it effectively”.
That sentence is the structural finding. Recognition of value and readiness to act are not the same thing. Most professional development conversations in education conflate them. Workshops that explain what AI is don’t build the muscles to actually use it well.
AI Literacy Among Librarians
The confidence numbers reveal something subtle. On the easier dimensions of AI engagement, librarians self-rated at moderate levels. The biggest gap was on troubleshooting AI tools, with 69.76% reporting low confidence. The other dimensions clustered together: collaborating on AI projects (40.16%), discussing AI integration (34.85%), and evaluating ethical implications (29.50%) all showed low-confidence rates.
The pattern points to a workforce that knows enough to recognize what they don’t know. That’s not a weakness. That’s the foundation for productive professional development. The dangerous workforce is the one that overestimates its capability. Lo’s respondents were calibrated. They named the gaps.
Training Follows the Hierarchy
One of the most striking subgroup patterns in the data is who gets AI training. Senior management had the highest training participation rate (47.27%). Support staff had the lowest (3.70%). Specialists fell in the middle.
That gap should bother anyone designing professional development. It means the people who actually deliver library services to students every day, the specialists and support staff, are the least likely to have received any AI training. The same thing happens in schools. Administrators get to attend AI conferences. Classroom teachers get a one-hour PD session. Lo’s data is the empirical version of a problem we already knew was there.
The Ethical Signal Is the Strongest in the Data
Across all the survey items, the strongest signal was on ethics and privacy. 74.34% of respondents rated addressing ethical and privacy concerns as urgent or extremely urgent. That’s stronger consensus than anything else in the survey. The qualitative responses sharpened the same theme. Librarians worry about implementation without consultation, about misuse and bias, and about libraries treating AI as a forgone conclusion when it should be a deliberate choice.
Lo notes that “these findings indicate the need for support systems, training, and resources to address readiness gaps, alongside rigorous discussion, and guidelines to navigate ethical and privacy issues as libraries explore the possibilities of AI integration”. Ethics has to come first, then training, then adoption.
The Framework That Foreshadowed ACRL
Buried in the conclusion is Lo’s seven-competency framework for AI literacy in libraries. It covers AI capabilities and limits, use cases, effective tool use, ethical evaluation, informed discussion, data privacy, and stakeholder impact.
This framework foreshadows much of what later appears in the 2025 ACRL competency framework, which Lo himself chaired. The 2024 paper is the empirical motivation; the 2025 ACRL document is the institutional response. Most surveys generate recommendations. Lo’s generated a profession-wide framework a year later.
What This Means for Teachers
The recognition-action gap Lo measures in librarians almost certainly exists in teachers too. Chiu’s (2025) editorial on the AI literacy versus competency split, which I’ve covered before, makes a related conceptual point: literacy is the compass, competency is the engine. There’s a difference between knowing what AI does and being able to use it for teaching.
Three things from this paper are worth taking seriously. First, the people in the room: the management-only training pattern Lo documented is exactly what to avoid. The content matters next: workshops can’t stop at “what is AI” without building troubleshooting and ethical evaluation skills. The hardest is measurement. Self-reported value isn’t readiness; capability has to be demonstrated.
Chee, Ahn, and Lee’s (2025) AI literacy competency framework, which I’ve covered before, makes a similar point about layering. Foundational concepts come first. Practical application is the next layer. Ethical integration runs throughout. Lo’s data shows what happens when those layers don’t get built systematically.
Where I’d Push the Argument
The paper has limits worth naming. The data was collected in 2023, when generative AI was new and few people in any profession had real fluency with it. We’re in 2026 now. Some of the literacy gaps Lo measured have probably narrowed as ChatGPT, Claude, and similar tools have matured into daily-use products. The training infrastructure has grown too, though unevenly.
The other limit is the self-report problem. Lo names this clearly. Self-rated AI literacy is not actual AI literacy. The respondents who feel most knowledgeable may know less than they think. Future research, as Lo suggests, needs objective assessments. Until then, the findings are best read as a description of professional perceptions, not professional capabilities.
References
- Association of College and Research Libraries. (2025). AI competencies for academic library workers. American Library Association. https://www.ala.org/sites/default/files/2025-10/acrl_ai_competencies.pdf
- Chee, H., Ahn, S., & Lee, J. (2025). A competency framework for AI literacy: Variations by different learner groups and an implied learning pathway. British Journal of Educational Technology, 56, 2146-2182. https://doi.org/10.1111/bjet.13556
- Chiu, T. K. F. (2025). AI literacy and competency: definitions, frameworks, development and future research directions. Interactive Learning Environments, 33(5), 3225–3229. https://www.tandfonline.com/doi/full/10.1080/10494820.2025.2514372
- Lo, L. S. (2024). Evaluating AI literacy in academic libraries: A survey study with a focus on U.S. employees. College & Research Libraries, 85(5). https://crl.acrl.org/index.php/crl/article/view/26409/34344