Why the room still needs a person
Teaching a language is not the same as knowing one. A professor runs a live room: opens a discussion in the target language, hears the hesitation in a student’s answer, and changes the task on the spot. A translation model can render a sentence cleanly. It cannot see that a student keeps dodging the subjunctive because being wrong out loud feels risky.
Two tasks carry most of the weight here. The first is delivering lectures and leading seminar discussion, usually partly in the language being taught, with the pace set by who is lost and who is bored. The second is evaluating student work — essays, oral exams, translations, literary analysis — and explaining in plain terms what to fix next. Feedback only works when the person giving it knows the student’s history and can be argued with.
There is also the institutional layer. Faculty advise students, plan and revise curricula, sit on committees, write recommendations, and sign off on grades that land on a transcript. Credit and accreditation rest on a named human judgment. That mix is what the still needs a human score is built to capture, and you can see how the whole scoring system fits together in our methodology.
What AI runs, what it assists, and what it leaves alone
Some pieces of the week can run with light supervision. Drafting vocabulary drills, generating practice dialogues and reading passages at a chosen level, and keeping attendance and grade records are the clearest cases. Our task split puts that share of task time in the AI-led group (12%).
A bigger block is assisted work. Preparing lecture materials, compiling bibliographies and reading lists, and building first-pass rubrics or quiz banks all move faster with a model, but a teacher still chooses, corrects and sequences. That assisted share is marked here too (53%). Can AI do it? scores 42 for this job, and the method behind that figure is on our coverage page.
The rest stays with people: live discussion in the target language, pronunciation and fluency coaching, judging whether a student’s reading of a text is defensible, advising on majors and study abroad, and departmental decisions. The share of task time in that group is shown above (35%). Nothing about this part is physical — no robotics are needed, which is why the timing here turns on software and institutions rather than hardware.
What the evidence does and does not show
There is no head-to-head test of AI against qualified postsecondary language faculty on this job’s real tasks. The evidence grade on this page (D) says exactly that, so we publish no parity number for it. Is it better than a person? is an open question here, and our quality parity page explains why an ungraded answer is better than a guessed one.
What would settle it is specific: blind grading studies where instructors and models mark the same student essays and oral exams, and term-length comparisons of language gain for sections taught with heavy AI support versus sections taught the usual way. Until that exists, claims that models already teach a language better than faculty are marketing, not measurement. If you want to see how general-purpose assistants answer the same question, our what the AIs say list collects their replies.
The labor market numbers are steadier ground. The US Bureau of Labor Statistics counts about 19,830 people in this occupation with median pay near $79,350, and projects employment change of roughly 0.2% from 2025 to 2035 (BLS, 2025). Flat is not the same as safe, and it is not the same as shrinking either. Enrollment in language programs matters more to headcount than any model release.
When the picture could shift
Most likely between 2034 and 2044 (8 in 10 of our scenarios). What that range actually measures is set out on our replacement year page.
Two things could pull the timing earlier. Cost is one: a year of classroom AI tooling sits far below the cost of a faculty line, so cash-tight departments have an obvious incentive to run larger sections with more automated practice and feedback. Enrollment is the other. If fewer students sign up for a language major, institutions may consolidate courses and lean on self-paced software rather than hire.
Two things hold it back. Accreditation and grading authority still require a responsible human instructor, which blocks any clean swap. And the core of the job is unscripted interaction — correcting a learner mid-sentence, reading confusion on a face, pushing a seminar past a shallow answer — which no current system handles end to end for a full term.
Good to know: the honest near-term story is task erosion, especially in material prep and first-draft grading, plus fewer openings for adjuncts and new PhDs.
How to stay needed
Lean into the tasks that stay human. Make live, high-pressure speaking practice the center of your courses, not an add-on. Own the part of assessment that requires judgment: oral exams, translation defense, and feedback on how a student argues about a text. Keep advising close — course planning, study abroad, graduate applications — because that is relationship work with institutional weight behind it.
Two skills pay off. First, assessment design that cannot be outsourced to a chatbot: in-class speaking, process work, drafts defended out loud. Second, fluent use of AI tools for prep and differentiation, so you set the standard in your department instead of reacting to it. Rank-and-file familiarity with these tools is now part of the job.
Nearby roles worth comparing: English Language and Literature Teachers, Postsecondary, Area, Ethnic, and Cultural Studies Teachers, Postsecondary, and Adult Basic Education and English as a Second Language Instructors. You can put any two of them side by side on our compare tool, see the wider postsecondary teachers family, or read how the rest of the education sector scores before you make a decision about graduate school or a department move.