When AI Guidance is Hard to Navigate

AI is changing research faster than shared norms can keep up. As researchers use AI to write, review, analyze, translate and assess scholarly work, the question is no longer whether AI belongs in research. It is how to use it responsibly, transparently and with the right safeguards in place.

This is a sponsored post from LIBER Gold Sponsor Frontiers, written by Simone Ragavooloo.

Reflecting the LIBER 2026 theme, The Power of Libraries in an Uncertain World, Simone Ragavooloo, Research Integrity Portfolio Manager at Frontiers, explores why research libraries have a vital role to play in helping researchers navigate AI guidance, understand responsible use and turn uncertainty into informed, ethical practice.

AI use is growing, but guidance is still hard to navigate

AI use in research is no longer hypothetical. It is now part of how researchers write, review, analyze, translate, and assess scholarly work. But while adoption is mainstream, shared norms for responsible use are still catching up.

Frontiers’ recent survey on AI adoption, barriers, and support highlighted a critical gap. Among active researchers surveyed, 77% reported using AI tools to support manuscript preparation and publication. At the same time, concerns about misuse are widespread, with 53% reporting that they had observed what they believed to be AI misuse by peers. Respondents also spoke of a lack of clarity around where the boundaries of responsible use lie.

If researchers are concerned about misuse, the logical next step is to seek trusted guidance on what responsible AI use looks like in practice. But the survey suggests that guidance is not always easy to find, interpret or apply.

When asked how they ensure best practice when using AI, researchers most commonly selected self-guided training materials on ethical AI use, at 35%, followed by guidance from their institution, at 31%, and feedback from peers or co-authors, at 30%. Only 16% reported using external policy pages from publishers, while 18% said they take no action to ensure best practice. Around one in five also identified lack of guidance or unclear rules as a barrier to responsible AI use.

These findings underpin one of the clear and most pressing calls to action: the need for a multi-stakeholder approach to AI literacy programs that help researchers understand responsible use, ethical risks, disclosure expectations, and the limits of AI tools and how to choose the right ones.

Fortunately, the sector is not starting from scratch. Quick mobilization means that practical guidance already exists, and the response from publishers and institutions has been swift. Wiley for example, has introduced author-facing AI guidelines, while Frontiers’ AI Playbook goes further, offering domain-specific, ready-to-apply guidance for researchers, reviewers, and editors. Crucially, it also calls on institutions to adapt this guidance locally, so they can better support researchers in navigating AI responsibly.

Early outreach suggests that practical guidance has the greatest impact when it is brought directly into institutional settings. Following a Frontiers seminar on responsible AI, the proportion of university participants who agreed or strongly agreed that they knew where to find responsible AI guidance rose from 33.3% to 83.3%. Confidence in identifying potential AI misuse or misconduct also increased, with agreement rising from 33.3% to 66.7%.

While based on a small sample, these findings point to an important lesson: guidance does not work simply because it exists. It works when researchers can find it, understand it and apply it to the decisions they face in practice.

The real challenge, then, is embedding responsible AI guidance into everyday research support. Even strong guidance has limited impact if it sits outside the places where researchers ask questions, make decisions and seek advice.

This is where librarians can make a real difference.

The growing role of librarians in AI literacy efforts

Librarians have long been central to scholarly communication. They support researchers with information literacy, open access, copyright, repositories, publishing choices, and responsible use of knowledge. In research institutions, they often act as trusted guides through a complex system of evidence, access, policy, and publication. This role is now becoming even more important.

If part of the challenge of AI literacy is helping researchers navigate information, make informed choices, understand policy expectations, assess evidence quality, and use tools responsibly, then AI literacy can be seen as a natural extension of this established role.

To be specific, Librarians can support AI literacy efforts in two connected ways.

First, they can help researchers evaluate AI-mediated information itself. As researchers increasingly use AI tools to generate, summarize, translate, or interpret information, they need to judge whether outputs are accurate, reliable, appropriately sourced, and fit for purpose. This builds directly on librarians’ established expertise in information literacy, source evaluation, evidence quality, and responsible reuse.

Second, librarians can help researchers navigate the growing body of guidance on responsible AI use. Researchers may ask: Can I use AI to help write a manuscript? Can I use it to summarize literature? Can I put unpublished data into a tool? Can I use AI during peer review? Do I need to disclose it? What does my institution, funder, journal, or publisher expect? Where can I find trusted and up-to-date guidance?

These are different questions, but they point to the same need: trusted support in making informed decisions. AI literacy is not only about knowing how to use tools. It is about knowing how to select them, evaluate their outputs, understand the rules around their use, and apply judgement in real research contexts.

In this sense, librarians can support the AI literacy movement by doing what they already do best.

From guidance to practice

For publishers and other stakeholders shaping AI expectations, guidance cannot stop at policies, principles or online resources. It has to reach the people helping researchers make decisions in practice.

That creates a clear opportunity for collaboration. Publishers can help set expectations for responsible AI use, while libraries can help make those expectations findable, understandable and usable. But this should be a partnership from the start, not a handover after guidance has already been written.

Libraries can test whether guidance works in real research contexts. They can show where language is unclear, where policies are difficult to apply, and where researchers need more support. They can also help localize guidance for different disciplines, research communities and institutional settings.

Frontiers’ AI Playbook was designed to help researchers, reviewers and editors navigate AI use responsibly. But for guidance to have real impact, it must be shaped and embedded with the communities that support researchers every day.

We invite libraries to help shape, test, localise and embed the AI Playbook. In an uncertain world, trust will not be built by guidance alone. It will be built by people working together to help researchers turn uncertainty into responsible action.

Libraries are already doing that work. They should be recognised as essential partners in what comes next.

To continue the conversation on responsible AI and the evolving role of research libraries, visit our LIBER 2026 page and book a conversation with the Frontiers team in Trondheim.