Editor’s note: Today’s post is by Dominique de Roo, Chief Strategy Officer at De Gruyter Brill.
When nearly 90% of academic librarians say AI is now recognized as a legitimate tool at their institution, it might seem like the debate has been settled. In our recent global survey of 311 librarians across 31 countries, that was indeed the headline figure. But legitimacy, it turns out, is not the same as enthusiasm — and the data beneath that number tells a far more complicated story. Acceptance at the institutional level does not translate into conviction at the personal level.
Indeed, across multiple regions, librarians described themselves as more cautious about AI than both administrators and academic staff. This is perhaps unsurprising: unlike administrators focused on deployment or researchers excited by new capabilities, librarians are uniquely positioned to see both the potential benefits and the practical drawbacks of the technology.
Regional Variations
In the United States, around 30% of responses indicated active resistance or non-use; in the UK and Ireland, nearly 40% of responses were blank or negative. Canada showed particularly high levels of skepticism. Respondents from the Middle East, Africa, India and the Asia-Pacific region tended to be more pragmatic — more often applying AI to specific, defined tasks. The pattern suggests that attitudes toward AI in academic libraries may be shaped as much by cultural and institutional context as by the technology itself.
While it’s true that AI may be viewed as “legitimate,” it’s far from universally loved. Understanding why that distinction exists tells us something important about how the technology may ultimately be adopted — and governed — within academia. Part of the explanation lies in how AI is currently entering universities. In many cases, the push is infrastructural: campus-wide licences for AI-enabled tools are increasingly bundled into existing enterprise software, particularly through large technology providers. This pattern, visible across the higher education sector more broadly, helps explain why institutional acceptance can advance faster than individual enthusiasm. For IT departments and leadership teams, these integrations can appear as upgrades to services and systems already in place. For publishers and content providers, this bundling dynamic raises an important question: if AI is arriving through the infrastructure rather than through the library, where does that leave the relationship between librarians and the scholarly content they curate?
The Library Caught in the Middle
Librarians, by contrast, tend to encounter AI from a different vantage point. In many universities, they are increasingly positioned between two forces. There are managers and technology teams pushing to implement AI tools rapidly across campuses. There are also researchers experimenting with AI to process large datasets, summarise literature or accelerate aspects of their work.
Both groups often turn to the library for guidance. This places librarians in a role that is not simply about end-use but about interpretation – asking questions about how these systems should be used, what risks they introduce and how they interact with established academic methods. One respondent captured the complexity of this position well: “It’s all over the place, from total ban to total embrace, at individual faculty level — we’re all resorting to ‘academic freedom’ to let faculty make their own rules for their students.” That positioning, however uncomfortable, gives libraries a form of influence that few other departments currently hold. In the absence of coherent policy from above, the library becomes de facto the place where norms around AI use are shaped, tested and communicated.
It is a role that sits at the intersection of guidance, governance and trust — and it is increasingly central to how institutions navigate AI adoption.
A Profession Built on Trust
Trust is a central theme in how librarians are approaching AI. Libraries have long positioned themselves as intermediaries in the scholarly ecosystem — curating collections, verifying information sources and helping researchers navigate an increasingly complex information landscape. The research shows that AI introduces new uncertainties into that environment. Many librarians remain uneasy about the ‘black box’ nature of large language models and other AI systems. These technologies can generate convincing outputs but often provide limited transparency about how those outputs are produced.
This uncertainty has practical implications. For example, some libraries are experimenting with AI-powered chatbots on their websites to assist patrons. Others have chosen not to deploy them at all, concerned that inaccuracies could undermine the library’s reputation for reliability. Survey respondents confirmed this concern is not merely theoretical: as one put it, “AI does hallucinate and invent articles that people then request, and it wastes our time — and theirs.”
In other words, the caution surrounding AI is not simply technological. It is reputational. If libraries are seen as reliable arbiters of information quality, then introducing tools that may produce unreliable outputs poses a difficult trade-off.
Divided Opinions within the Profession
Another striking feature of the survey, conducted by Gold Leaf, was the diversity of attitudes toward AI within the librarian community itself. Some respondents expressed enthusiasm about the possibilities AI might unlock — particularly in areas such as metadata generation, discovery tools, and supporting researchers working with large datasets. Others were openly skeptical — even hostile. As one respondent described the range: “It is truly a mixed bag, from embracing in all forms, to acceptance and exploration, to willing to learn/explore, to full stop/abhorrence.”
The research reveals that individual institutional circumstances appear to play a significant role in shaping these attitudes. In libraries where budget cuts have led to redundancies and librarians have absorbed more routine tasks, AI is often welcomed as a practical aid. But where cuts have been more severe, that pragmatism moves closer to anxiety and a fear that AI will erode core professional responsibilities.
Between these poles lies a substantial group who are neither enthusiastic nor resistant, but cautious — aware of the technology’s potential while also wary of its risks. This middle position may ultimately prove the most influential. Rather than embracing or rejecting AI outright, many librarians appear to be asking a different set of questions: how can AI be integrated into academic practice without undermining the principles that scholarship depends on?
The Governance Gap
One of the most revealing findings in the survey is that librarians are still grappling with these questions. Fewer than half of respondents (47%) reported that their library has developed formal guidelines for AI use. That said, many of those without guidelines reported that work was underway, or that AI was considered sufficiently covered by existing ethics policies. The more revealing issue may be one of ownership: in many institutions — particularly in Europe, but elsewhere, too — responsibility for AI governance does not sit with the library at all, falling instead to IT departments, research offices or senior leadership. Among librarians themselves, views on this are divided. Some feel strongly that it should be their remit; others are relieved it’s not; while others still have simply absorbed the responsibility by default. One respondent described the resulting fragmentation in stark terms: “There is no point person, no centralized unit spearheading AI efforts — just work happening in different pockets across campus.” This uncertainty reinforces the idea that AI adoption in universities is still in its early stages — even if the technology itself is advancing rapidly.
From Custodians to Stewards of AI-Mediated Scholarship
For many librarians, the rise of AI represents not only a challenge but also an opportunity. The profession has spent decades evolving from managing physical collections to supporting digital scholarship, data management and open access publishing. AI may represent the next phase of that development.
If libraries become (or are beginning to become) the institutions responsible for interrogating how AI interacts with scholarly methods — asking questions about provenance, transparency, attribution and integrity — then their role in the academic ecosystem could become even more central.
The survey data suggests many librarians recognize this. As one respondent put it: “The consensus is that AI is a transformative tool that requires guidance, not rejection.” In this sense, the current caution surrounding AI may reflect not resistance but responsibility.
What This Means for Scholarly Publishing
For academic publishers, these dynamics are significant. Libraries have long served as key intermediaries between publishers and researchers. As AI disrupts how knowledge is discovered, interpreted and produced, librarians may play an equally important role in shaping how these technologies are integrated into scholarly workflows.
Their caution signals that adoption will not simply be a matter of technological capability. Questions of trust, governance and academic integrity will remain central. Publishers therefore face many of the same questions librarians are already facing: how should AI tools be integrated into scholarly infrastructures and what safeguards are necessary to maintain confidence in the research record? What the research suggests librarians most want from publishers is not necessarily direct involvement in institutional governance, but transparency. Clear communication about how publishers themselves use AI, what protections are applied to content, and above all what steps are being taken to preserve research integrity in an environment where the authenticity of scholarly work can no longer be taken for granted.
Legitimate — But Still Contested
The survey’s headline statistic — that nearly 90% of librarians recognize AI as a legitimate institutional tool — captures only part of the story. Legitimacy reflects acceptance that the technology is here. It does not necessarily signal confidence in how it will be used.
Across the global librarian community, AI is simultaneously viewed as an opportunity, a risk and an open question. And in between those perspectives lies something important: the future of AI in academia will not be determined solely by how quickly it is deployed but by how carefully it is governed. Libraries may prove central to that process. The question is whether institutions, publishers and policymakers will recognise this before the decisions that matter most have already been made.
