Why this job holds on to people
Will AI replace library science teachers? Not on the evidence in front of us. The work is split between content that software can draft and judgment calls that a person has to own. Writing a lecture on metadata standards is one kind of task. Deciding whether a graduate student is ready to run a reference desk in a public library is another. Only one of those moves cleanly to a model.
Much of the job is teaching adults who will soon advise other people on how to find and trust information. Faculty lead seminars, run cataloging and reference exercises, grade research projects, and advise students through practicum placements. They also sit on curriculum and accreditation committees, where a program’s standing depends on named academics defending their choices. Those duties carry responsibility, and responsibility is hard to hand to a tool.
There is a second reason the role is stable in an odd way. AI is now part of what library science faculty teach. Information literacy courses have to cover how generative systems retrieve, summarize and fabricate. That is new material someone has to design, test and update each term.
What AI does, what it assists with, and what stays human
AI handles the production work around a course. It drafts slide decks and discussion prompts, summarizes assigned readings, builds quiz banks from a syllabus, and formats course documents. Share of task time where AI can do the work: 7%. None of that requires a hardware investment, because the job has no physical component.
A larger block is assistance rather than substitution. Models can suggest reading lists, generate practice records for a cataloging exercise, and give first-pass comments on a student draft that the instructor then corrects. Share where AI assists a person: 54%. The instructor still sets the standard and signs the grade.
Then there is the part that stays with a person: live seminar discussion, mentoring and reference writing, program review, and judging readiness for field placement. Share of task time that still needs a human: 39%. Our task-time measure is explained on the coverage method page, which reports a coverage figure of 39 out of 100 for this occupation.
What the evidence actually shows
No published study has tested an AI system against library science faculty at their own work. That is why the evidence grade here is D, and why we publish no quality figure for this job. A grade at that level means not measured, not measured and found wanting. The two are easy to confuse and very different.
What would settle it is specific. A graded comparison of model-written and instructor-written feedback on student cataloging, metadata and reference assignments, scored blind by independent faculty, would give a real number. So would a course-level trial in an online MLIS program comparing learning outcomes with and without AI-led instruction. Until something like that is published, the honest answer is that the question is open. How grading works is set out on the quality parity method page.
The labor market data is steadier. BLS counts about 3,630 people in this occupation, with median pay of $80,340 and projected employment change of 2.7% between 2025 and 2035 (BLS, 2025). That is slow growth, not decline. The pressure on this field has more to do with enrollment in library science programs than with software.
When the picture could shift
Most likely between 2034 and 2045 (8 in 10 of our scenarios). What that range means is explained on the replacement-year method page.
Two things could pull the date earlier. Online graduate programs already lean on automated grading and asynchronous content, and cheap course production makes larger sections easier to run with fewer instructors. Budget pressure on small programs points the same way: departments merge, adjunct hours get cut, and tenure lines go unfilled before any task is fully handed over.
Two things push the other way. Accreditation and program review expect named faculty to take responsibility for curriculum and student assessment, and that expectation changes slowly. And trust is still thin: a model that invents a citation is a poor teacher for a field built on verifying sources. Because the role needs no robotics, nothing is waiting on hardware. The brake is institutional, not mechanical.
How to stay needed
Lean into the work that sits in the human column. Lead the live seminar rather than the recorded lecture. Keep ownership of practicum supervision and student assessment, where someone has to vouch for a person’s readiness. Take a seat on curriculum and accreditation work, because that is where programs decide what gets taught next.
Two skills matter more each year. First, teaching AI literacy well: how retrieval systems rank, where hallucinated citations come from, and how to check a generated answer against a primary source. Second, assessment design that survives generative tools, including oral defenses, in-class cataloging work and process-based grading.
What to do: rebuild one assignment this term so a student has to show their search path and sources, not just the finished answer.
If you are weighing adjacent paths, the closest work sits nearby: Education Teachers, Postsecondary, Communications Teachers, Postsecondary and Librarians and Media Collections Specialists. You can put any two of them side by side on the job comparison tool, or browse the wider postsecondary teachers family and the education sector pages. Our scoring is documented at how we score jobs, and the jobs that mostly need a person list shows where teaching roles land against everything else.