Four Quick Charts on AI Implementation in Academic Libraries

What our micro-course participants had to say about AI in libraries

Posted on August 12, 2026 in Blog Posts

Here at LibTech Insights, we know that only thing better than a “Which Sex in the City character are you?” test is a poll about AI implementation at your library. It’s easy enough to talk about “AI in libraries,” identify current challenges and opportunities, and speculate about future development, but it’s much harder to get a clear on-the-ground picture of the state of AI in academic libraries.

As a part of our free AI literacy micro-course we recently launched with Clarivate, we asked participants to answer a series of questions about AI implementation at their library. This survey was a part of the first module, “Connecting Libraries to the AI Ecosystem.” The full module also includes an essay on library leadership in university AI governance, a star-studded roundtable recording, and a series of excellent readings on the topic. But for this post, we wanted to take some of the survey results and discuss the surprises (and unsurprises) we found.

A few caveats before we get started. Survey respondents were anonymous, so we have only the raw data. What this means is that some respondents might not be academic librarians, though most likely are, and some respondents might be from the same institution, leading to duplicate data points. So, while we admit that the findings aren’t altogether rigorous, we maintain that they are still an interesting glimpse into the state of AI in academic libraries.

🔥 Register for AI Essentials for Academic Libraries: Beyond the Basics

---

At first glance, the responses to this survey question about where their library currently stands in terms of implementing AI tools would seem to tell a straightforward story. Looking at the bar graph, it looks like libraries are steadily marching toward active implementation. While that might be the case, the data have a few wrinkles.

Perhaps most surprising is the size of the minority of respondents reporting that their libraries neither had plans to implement AI nor were actively pursuing it—a combined 21-percent chunk. Though overshadowed by the 79 percent either pursuing or implementing AI in some capacity, it nonetheless represents a solid bloc.

These data also take on a different appearance if we take out the glut of respondents who report that their libraries are in the “exploration and evaluation” stage (30 percent). The actual breakdown becomes far more undecided: 21 percent aren’t seeking to implement AI, 30 percent are evaluating whether to do so, and 49 percent are actively deploying it to some degree. Or, viewed differently, 49 percent of libraries are actively deploying AI tools in collections and services, and 51 percent aren’t.

The point is, when we talk about “AI in libraries,” we’re looking at a mixed, if not divided, landscape.

Although Anthropic’s Claude seems to be enjoying a lot of online popularity lately, it lost out big in our poll asking respondents which general-use AI model they use for research. Unsurprisingly, Microsoft Copilot, given many universities’ investment in the Microsoft product ecosystem, won out at 24 percent, while OpenAI’s ChatGPT came in a close second at 23 percent. Perhaps given the question’s focus on research, Gemini, as an extension of Google, also ranked highly.

No real surprises, but these results show what AI companies know all too well: no one has truly captured the market.

---

Now live: A Pedagogical Road Map for Teaching AI Literacy

The second installment of Choice and Clarivate’s free, three-part micro-course, “AI Essentials for Academic Libraries: Beyond the Basics,” is now live! In this module, written by and for librarians, we cover pedagogical frameworks for AI, ready-to-go library activities for engaging learners, and many resources for further learning and discussion. Register now to access the first two modules of the course.

REGISTER NOW

---

Respondents were allowed to choose up to three responses when asked what their biggest concerns were with applying AI technology to research platforms and databases. The responses showed a big, but perhaps understandable, divide, given the question’s focus on research platforms: respondents expressed far greater concerns about research-related problems (misinformation, copyright, academic integrity) than broader social concerns (environmental costs) or financial concerns. “Bias and discrimination in AI outputs,” though it arguably falls within research-related problems, was in the lower half.

The other notable exception is that privacy ranked among the top three concerns. In my experience, most people are resigned to data collection, even if they oppose it in theory. But perhaps, given AI companies’ open thirst for data, respondents are more actively concerned with privacy. The United States is also undergoing something of a public reckoning over surveillance due to Flock cameras, not to mention widespread opposition to the data center buildout, so the concern for privacy is, in general, on the rise.

AI companies are hungry for data, and libraries, particularly in their special collections and archives, have a lot of unique data. Indeed, earlier this summer, Dr. Leo Lo gave a webinar on this very issue and offered a framework for libraries considering training requests from AI companies. If your library hasn’t received such a request, rest assured that it’s only a matter of time. But right now, the landscape seems pretty decided: 64 percent of respondents reported that their libraries aren’t allowing AI training and aren’t planning to do so.

Of course, this means that 36 percent of libraries represented by the respondents are either allowing it or considering—perhaps a higher number than one might suspect. Given that half of this segment falls into the “not currently, but we have plans to” stage, Dr. Lo’s work becomes more important: libraries need to ensure that their agreements with any institution, including their own universities, to train LLMs on their collections align with their values.

---

This post highlighted only a few of the survey questions we asked in this micro-course module. Want to see the full results? Sign up for the micro-course; it’s not only free, but it’s also fun and informative. The second module, on AI pedagogy, is now up and comes complete with an activity bank, submitted by librarians and college educators, filled with exercises they use in their AI literacy sessions.

Related Posts