06/24/2026
In 1894, librarians faced a crisis.
The open-access library had just arrived. For the first time, ordinary people could walk the stacks and choose their own books. It sounded like progress. What actually happened was that a blacksmith looking for information on tempering iron came home with a 500-page theological sermon called The Fiery Furnace because the keywords matched.
Sound familiar?
A generation of librarians including Samuel Swett Green and Melvil Dewey responded by building something the open stacks alone could not provide: structured access, classification systems, and the reference interview. They understood that access to a large collection and the ability to retrieve what you actually need are two entirely different things.
Are we back in 1894?
The most sophisticated information system ever built is sitting in front of millions of people, and most of them are typing in a vague question, hoping for a useful answer, and pasting in whatever comes back. The "hope and paste" method, as one writer recently called it. The result is surface-level retrieval from a system that contains extraordinary depth.
Here is the contrarian view: AI has not made librarians less relevant. It has made the core skill of librarianship more valuable than it has ever been. Structured query design, metadata discipline, understanding how a collection is organized and what it actually contains; these are not legacy competencies. They are exactly what separates useful output from a sophisticated-sounding guess.
And before any of that matters, the collection has to exist in a form the system can actually use. A digitized document with good capture quality and proper metadata can be retrieved. A poorly scanned image sitting in an unindexed folder cannot be, regardless of how well you prompt.
