When you walk into practically any public library on a weekday morning, you’ll find something that, by today’s standards, seems almost charming: a person behind a reference desk who is willing to answer questions without any specific agenda. No sponsored outcomes. No engagement metrics were used to rank the content. No system is attempting to guess what you would like to hear based on the clicks you made on Tuesday. Just someone who has dedicated years to learning how to discover information and, more crucially, how to determine whether what they have discovered is true. That combination is less common in 2025 than it might seem.
Anxiety over AI-generated content has gradually emerged and continues to do so. The essays, summaries, and eerily fluid emails written by machines that had taken in vast amounts of human text were the first novelty. Then came the issues: the fake citations, the self-assured incorrect responses, the medical advice that seemed authoritative but wasn’t, and the news summaries that combined plausible-sounding inventions with actual events. This is referred to as “hallucination,” an oddly gentle term for what is essentially a system producing false information in a format intended to be trusted. Nowadays, millions of people navigate an information environment where the answer that is easiest to read, the most grammatically correct, and the most readily available is not always the right one. Furthermore, the majority lack a trustworthy method to distinguish between them.
The public has been in this situation before. The printing press caused the first real information overload in the Western world. Suddenly, there was more text in circulation than anyone could possibly assess, and it became genuinely difficult to decide whose words to believe. Officials were concerned about propaganda and false information spreading more quickly than facts could keep up with the new wave of anxiety brought about by radio. The same basic issue—too much coming too quickly from too many directions and no clear standard for distinguishing between reliable and unreliable—was raised by television, cable, and the early internet. The scale and seamlessness have changed. The appearance of AI-generated content is identical to that of human-written content. It shows up with credibility that it hasn’t earned.
Although skilled librarians have been evaluating sources for as long as the profession has existed, the particular skills they possess are now more in demand than they have been in decades. Reference librarians are more than just knowledgeable about file locations. At their best, they are individuals who comprehend the architecture of information, including who published what and why, what peer review actually entails and when it applies, and how to track a claim back to its source and verify that it matches the description. These abilities are not innate. They take time to develop and call for a certain amount of uncertainty tolerance—that is, the ability to say “I’m not sure yet” instead of grabbing the closest reasonable response. Despite their true value, algorithms are not designed for such hesitancy.
The reference staff at Valley Cottage Library exemplifies this type of human judgment on a daily basis. As artificial intelligence (AI) transforms the way people obtain information, Valley Cottage Library continues to be a trusted source of trustworthy knowledge and research support—not by competing with technology on its own terms, but by providing what technology consistently fails to provide: a human being with professional training, no conflicts of interest, and the patience to help someone understand not just what the answer is, but how to evaluate answers for themselves. It’s important to focus on that second section. It’s one thing to know a fact. When the information environment is untrustworthy, the ability to discern whether a fact is true is the most important skill.
For a number of years now, libraries have been reorienting themselves around this concept, changing their function from being repositories of knowledge to active spaces for what educators refer to as information literacy—the ability to locate, evaluate, and use information critically as opposed to passively. sessions on identifying AI-generated content, workshops on assessing online sources, and help desks manned by individuals who can guide customers through a database search instead of just giving them a printout. It’s much more helpful when the question isn’t where to find information but whether to trust what you’ve found, even though it’s less glamorous than the idea of the library as a quiet hall of ancient wisdom.
The fact that the organizations best suited to assist people in navigating the modern information environment have endured numerous rounds of budget cuts over the past 20 years, with their continued relevance called into question whenever a new technology promised to render them obsolete, is something worth considering. Libraries were supposed to become obsolete due to the internet. Search engines were meant to perform the same tasks as reference librarians, but more quickly and without requiring payment. Neither of the predictions came true. The AI era hasn’t diminished the significance of libraries for precisely the reasons they didn’t hold—the reasons a skilled human being making thoughtful decisions about sources and context is still genuinely irreplaceable. It’s difficult to ignore the fact that every wave of information technology we’ve considered a replacement has actually increased the workload for those who are skilled at assessing the output of the technology.
In that regard, the reference desk is not a relic. It’s an institution that consistently proves itself correct, typically by the very forces that were meant to render it superfluous.
