Librarians can’t keep up with bad AI

Generative artificial intelligence continues to have a problem with hallucinations. Although many answers to user queries are largely accurate, programs like ChatGPT, Google Gemini, and Microsoft Copilot still tend to offer made-up information and facts. As serious as this may seem in itself, the problem is further complicated by the tendency of these AI programs to produce seemingly reputable, but entirely imaginary, sources. But as annoying as this may seem to millions of users, it becomes a major problem for the people who are trusted to provide reliable, real information: librarians.

“For our staff, it’s much more difficult to prove that a single document doesn’t exist,” said Sarah Falls, research librarian at the Library of Virginia. Scientific American.

Falls estimates that about 15% of all reference questions his staff receives are written by generative AI, some of which include quotes and imaginary sources. This increased burden placed on librarians and institutions is so heavy that even organizations like the International Committee of the Red Cross are alerting the population to the problem.

“A specific risk is that generative AI tools will always produce an answer, even when historical sources are incomplete or silent,” the ICRC warned in a public advisory earlier this month. “Because their goal is to generate content, they cannot indicate that no information exists; instead, they will invent details that seem plausible but have no basis in the record. »

Instead of asking a program like ChatGPT for a list of ICRC reports, the organization suggests you go directly to its publicly available information catalog and scientific archives. The same strategy should be extended to any institution. Unfortunately, until more people understand the fallibility of generative AI, the burden will remain on human archivists.

“We will probably also let our users know that we need to limit the time we spend verifying information,” Falls warned.

There’s a good reason why librarians have remained an integral part of societies for thousands of years. Unlike generative AI, they are trained to think critically, search for answers, and most importantly, admit mistakes.

PopSci Outdoor Gift Guide 2025

20+ Editor-Approved Gifts for the Hikers on Your List