Why do we automatically respond to new technologies by wanting to create more? The Mexican writer Juan Villoro observes that our relationship with technology has evolved into one of dependence in which failure is defined by “the fear of missing out”. This is the fear that pushes us towards production fed by our tendency towards addition and complexity, a phenomenon well documented. When solving problems, we tend to add rather than subtract. Faced with the challenge that generative AI presents to learning and research in higher education, the same instinct has shaped our response. We are compelled to create more frameworks and competencies rather than pause to consider what we already have. When a new technology appears, we must, of course, help students navigate unfamiliar interfaces and terminology. Existing frameworks need translation, new terms must be incorporated, and new connections must be made explicit. But a good translation remains faithful to the original message. It presents information in a new language without inventing an entirely new text. What troubles me is not adaptation itself. Rather, it is the rhetoric of radical novelty that has developed around it; the belief that generative AI has severed us from foundational frameworks, that what worked before no longer applies. This rhetoric does not clarify, but obscures.

There are, to be fair, serious versions of the argument that something has genuinely shifted. Olof Sundin, for one, makes a careful case that AI-infused search breaks the link between content and its sources, and that our field may need to rethink core concepts like evaluation. This is a claim worth taking seriously. But it is a long way from the reflexive sense that our accumulated practice has become obsolete precisely when it is most needed — a story that tells us, in the same breath, that librarians are best placed to lead on AI policy and that everything we know is out of date.

It is perhaps counterintuitive, then, that in a time when the pace of technological change is so rampant, we might find some inspiration in the writings of the fourteenth-century English friar, William of Ockham.  Ockham’s Razor is often presented in the maxim “entities should not be multiplied beyond necessity”. Though, as Eliot Sober points out in Ockham’s Razors: A User’s Manual, that wording never actually appears in Ockham’s own work. What Ockham wrote is simpler: “it is futile to do more with what can be done with fewer.” Sober also reminds us that the razor has a positive side as well as a negative one: do not add something if you do not need it but do add it if it is essential. That is the thought worth carrying into our response to AI. It is what can keep us steady amid constant updates, inflated claims, and the pressure to add more just to stay relevant.

In practical terms, a new iteration of Ockham’s Razor for academic librarians might be expressed as: “find the familiar; the novelty will fall away”. By this, I mean, it is essential to look past the marketing and new interfaces of an AI tool and see the research process beneath it: what is the tool actually trying to do?; is it for literature searching, critical reading, note-making or writing? Focus on the function, and it will become apparent that the questions we have been teaching students to ask about sources, authority, and coherence haven’t become obsolete; they have become more important than ever.

Take, for example, large language models that invent false citations. When we teach students to check citations, we are not teaching something new. Yes, the interface has changed, but we have always needed to remain critical of sources even when we were searching in card catalogues and pulling books off the shelves.  Check your sources, trace the claims, and evaluate the evidence before you use it.  This is the foundational premise of academic research. Take also natural-language search, where students type a question, rather than a list of keywords, and receive an answer in a sentence. Leo Lo has named this shift the “answer economy,” and he is right that students increasingly receive a packaged answer where they once weighed sources. However, where I would disagree is, that from the student’s point of view, this is less a sudden rupture than the latest development in a longer process. As Olof Sundin observes, the move from a list of links to a ready-made answer was already underway on the open internet, through Google’s knowledge panels and featured snippets, well before generative AI arrived. Students have long received pre-packaged answers in lectures, textbooks, encyclopaedia entries, and increasingly from online search itself. What generative AI changes is not the basic pedagogical task but its urgency: students must still question answers rather than simply receive them.

In this age of generative AI, we must resist the urge to keep pace simply to remain relevant. As I tell faculty in my workshops, librarians are rather good at handling our professional existential crises. We survived the arrival of automation and the arrival of the internet. AI literacy is, at its core, information literacy. And when we feel the urge to build a new framework in response to the latest tool, let us pause first, and find the familiar. Perhaps the novelty will fall away.