Short answer: no, and the projections separate the profession cleanly into two halves that are moving in opposite directions. Librarians and library media specialists are projected to grow 3 percent between 2025 and 2035, adding about 3,600 positions to a base of 142,300 with 2025 median pay of $68,270. Library technicians and assistants are projected to decline 6 percent over the same period, losing about 10,400 positions. Same buildings, same institutions, opposite trajectories. The split tells you exactly which parts of library work are being automated: circulation, cataloguing support and processing, not selection, instruction, or the judgement about what is trustworthy.
Why the Two Halves Diverge
Library technician and assistant roles are built around handling material and transactions: checking items in and out, shelving, processing new acquisitions, basic cataloguing, and helping people find a specific known item. Self-checkout, RFID, automated sorting and shared cataloguing records have been eroding those tasks for two decades, and the 6 percent projected decline is the continuation of a long trend rather than a new AI effect.
Librarian roles are built around judgement: choosing what to acquire with a fixed budget, teaching people how to evaluate sources, designing services for a community, handling reference questions that are genuinely ambiguous, managing licences and privacy, and running programmes. None of that is transactional.
There is also a factor specific to this profession. When information becomes abundant and unreliable at the same time, the value of someone who can judge quality goes up, not down. That is exactly the environment generative systems have created.
BLS published AI exposure categories alongside the 2025-35 projections, sorting occupations into Low, Moderate, High and Very high relative exposure, and states directly that exposure “does not imply job loss, productivity gains, automation probability, or wage effects.” Library work involves language and structured retrieval, so it registers on exposure measures. The professional half is still projected to grow.
What Changed About the Reference Interaction
The classic argument against libraries was always “why not just search?” That argument got stronger, and it also revealed its own limitation faster than expected.
A model answers a well-formed question quickly and often correctly. It also produces confident citations that do not exist, misses paywalled or offline material entirely, cannot tell the user what a database does not contain, and has no view about whether the source it drew on is credible. Every one of those failures is a reference librarian’s core competence.
The observable effect in academic and public libraries is a change in the questions arriving at the desk. Simple factual questions have largely gone. What remains is harder: users arriving with a plausible-looking answer that turns out to be fabricated, users who need material that is not indexed anywhere useful, and users who need help deciding whether a source can be trusted at all.
| Task | Automation effect | What the librarian now does |
|---|---|---|
| Known-item lookup | Largely automated | Almost nothing, correctly |
| Ready reference facts | Automated | Verifying when the stakes are high |
| Database searching | Assisted | Choosing the right database and constructing a defensible strategy |
| Source evaluation | Not automated | Teaching it, and doing it for high-stakes questions |
| Verifying a fabricated citation | New work | Tracing what the user was actually given |
| Collection development | Not automated | Spending a fixed budget against community need |
| Privacy and licensing | Not automated | Negotiating terms, protecting patron records |
| Programming and outreach | Not automated | Designing services people actually use |
Notice that one row is new work created by the technology. Chasing citations that do not exist has become a recognisable category of reference question, and it is not a task any system solves for the user.
A reference interaction that could not be automated
A graduate student arrives with a reading list of eleven references produced by a chatbot for a literature review. Four are real and correctly cited. Three are real papers with the wrong authors or the wrong journal. Four do not exist at all, though two of those combine the title of one real paper with the authors of another.
Untangling that takes a person who knows how the literature is organised in that discipline, which databases index it, and what a plausible-but-wrong citation looks like. It also takes a conversation, because the student’s real problem is not the eleven references. It is that they do not yet know how to tell a real citation from a convincing one, and nobody has taught them.
The librarian’s output from that half hour is not a corrected list. It is a student who can now do the check themselves. That is teaching, it is the highest-value thing the profession does, and demand for it has risen sharply rather than fallen.
What to Know Before You Draw Conclusions
Libraries are funded politically, not technologically. The main threat to library employment over the next decade is budget decisions by municipalities, school districts and universities. Automation is a much smaller variable than funding, and conflating the two leads people to prepare for the wrong problem.
Information literacy demand is rising sharply. Schools and universities need someone to teach students how to evaluate machine-generated material. Librarians are the profession with the existing expertise and mandate.
Privacy expertise is becoming a differentiator. Libraries have an unusually strong professional tradition on patron privacy, and that tradition is directly relevant as institutions adopt tools that log everything. This is an area where librarians have credibility that others in the institution do not.
Special libraries are a separate market. Law firms, hospitals, corporations and government agencies employ information professionals where the cost of a wrong answer is high. That segment values verification most, which makes it the most resistant part of the profession.
The technician decline is real and should be named. Anyone in a circulation or processing role should treat the 6 percent projection seriously and plan accordingly, rather than being reassured by numbers describing a different job.
One further point is worth making plainly, because it is often left implicit. The professional argument for libraries has never rested on scarcity of information. It rested on the idea that a community benefits from an institution whose only interest is helping people find reliable material, with no commercial incentive attached to what they find. In an environment where most retrieval sits inside advertising or subscription businesses, that neutrality is a stronger differentiator than it has been at any point in the last thirty years, and it is worth saying out loud in budget conversations rather than assuming everyone already understands it.
Where the Work Is Growing
- Instruction and information literacy. Teaching source evaluation is the single largest growth area, particularly in academic and school libraries.
- Data and research support. Data management plans, repositories, research data curation and scholarly communication support are established, funded roles in academic libraries.
- Digital collections and metadata strategy. Automated cataloguing raised the value of people who decide what the metadata should describe and how collections connect.
- Community and social services roles. Public libraries increasingly function as the last open civic space, with staff supporting job seeking, benefits access and digital inclusion. None of that is automatable.
- Archives and special collections. Unique, unpublished and physical material sits outside any training corpus, and description of it remains expert work.
- Systems and discovery. Someone has to configure, evaluate and troubleshoot the discovery layer, and increasingly to assess the AI features vendors are adding to it.
A Practical Framework for Library Careers
- Library technician or assistant. The most exposed group. The strongest moves are toward public service and programming, toward systems work, or toward a professional qualification if it is realistic for you. Set a timeline rather than waiting.
- Public librarian. Your position is tied to community value. The services that protect budgets are the visibly human ones: programming, digital inclusion, and support for people who cannot get help anywhere else.
- Academic librarian. Instruction and research data support are growing. Being the person who can teach students to evaluate machine-generated sources is currently the highest-demand capability in the sector.
- Considering the profession. Enter through a specialism rather than generally. Data curation, health sciences, law and archives all have clearer demand than general reference.
The test that applies across all four: how much of your week is transaction handling versus judgement, teaching or relationship work? That ratio is your exposure, and every growth area above sits on the judgement side.
Building the Literacy You Are Being Asked to Teach
There is an awkward gap in many institutions right now. Librarians are being asked to teach students and staff how to evaluate machine-generated information, and many have not used these systems enough to describe their failure modes precisely.
Closing that gap is the highest-leverage professional development available in this field. It means understanding how these systems produce text, why fabricated citations look so convincing, what retrieval-augmented approaches change and what they do not, and how to design an exercise that shows a student the failure rather than telling them about it. That is a short, structured course of study rather than a technical specialism, and doing it deliberately puts you in a position to lead your institution’s response. A certificate alongside applied teaching practice makes that capability visible when a committee is deciding who writes the policy. If you want a structured route in, explore Coursiv AI lessons and check current plan details on the official site.
FAQ
Are librarian jobs disappearing?
Can AI answer reference questions?
Which library roles are safest?
Do libraries still need physical space?
What should a librarian learn now?
Your Next Step
Take one instruction session you already deliver and rebuild it around a live demonstration: ask a model a discipline-specific question, then trace the citations it gives you in front of the class. Students remember watching a plausible reference dissolve far better than they remember a slide about evaluation criteria, and the exercise doubles as evidence for your institution that this teaching needs a professional behind it.