Why the seminar room keeps a person in it
Teaching literature at a college is really two jobs. One is producing material: syllabi, reading lists, lecture notes, quiz items, written comments on drafts. The other is judging what a particular student understood and pushing that student further in real time. Language models are good at the first job. The second is where the work stays.
Grading is where the pressure is most obvious. A rubric is written down, so a model can apply it to a stack of essays quickly and consistently. But an English course is not mainly about the grade. It is about whether a student can build an argument from evidence in a text, defend it when challenged, and revise it after being told why it does not hold. That loop runs through conversation, and the person running it has to know the student.
Discussion sections make the same point. A seminar depends on reading the room: who has not done the reading, who is close to a real insight, which half-formed comment is worth stopping on. Add the advising load, thesis supervision, letters of recommendation, and the judgment calls about whether a submitted essay is a student’s own work, and a large slice of the week is contact, not content.
What AI does, what it assists with, and what stays with faculty
The routine production work is where AI already carries weight: drafting discussion prompts, summarizing secondary reading, generating practice questions, and making a first pass at sentence-level comments on a draft. None of that requires a body in a room. Task time in that group: 16%.
A larger band of the job is assisted rather than handed over. Lecture prep, syllabus revision, rubric design, feedback on structure and evidence, and keeping up with scholarship all go faster with a model in the loop, but a professor still decides what counts as a good reading and what the course is for. Task time in the assisted group: 47%.
What is left is the contact work: leading discussion, assessing original argument in person, mentoring revision, advising majors, serving on committees and in department governance. Task time that stays with a person: 37%. Taken together, our coverage score — can AI do it — comes out at 42 out of 100, and you can read how that figure is built on the coverage method page.
What has actually been tested
Not much, directly. Our evidence grade for this job is D, which means there is no published head-to-head test of an AI system against a qualified postsecondary English instructor on this job’s core work. So we publish no parity number here, and anyone quoting one for English faculty is guessing.
What would settle it is specific. Blind grading trials in which faculty and models score the same student essays against the same rubric, with a check on whether students who got machine feedback actually revised better. Controlled comparisons of discussion sections led by an instructor and by a tutoring system, measured on later written work rather than on satisfaction. Until that exists, the honest position is uncertainty about quality, not confidence in either direction. The quality parity method explains why a D grade never gets a score, and the full approach sits on our methodology page.
When the picture could shift
Most likely between 2034 and 2044 (8 in 10 of our scenarios). What that window measures is set out on the replacement year method page.
Two things could pull it earlier. Budget pressure is the first: tools to handle feedback and course materials run at a fraction of the cost of the faculty hours they touch, as the cost panel above shows, and nothing physical is needed, so adoption has no hardware barrier. The second is assessment drift. If written take-home essays keep losing credibility as evidence of learning, departments may redesign courses around automated practice plus a few proctored checks, which changes how many instructor hours a department buys.
Two things hold it back. Accreditation and faculty governance move slowly, and credit-bearing instruction is tied to credentialed people in ways no model satisfies on its own. And the part students pay for — being known, argued with, and recommended — is the part that does not transfer. On headcount, the Bureau of Labor Statistics counts about 57,720 US postsecondary English language and literature teachers at a median wage of $78,760, with employment projected to hold flat from 2025 to 2035 (BLS, 2025). Flat demand usually hurts the newest entrants first; our entry-level tracker follows that signal.
What to do: move more of your assessment weight onto in-class writing, oral defense of an argument, and graded revision, where the evidence of learning is produced in front of you.
How to stay needed in an English department
Lean into the three tasks a model cannot finish for you: running discussion that builds on what students actually said, assessing original argument and coaching the revision that follows, and advising — thesis supervision, recommendations, the long conversation about what a student is for.
Two skills raise your value quickly. First, assessment design: writing courses whose graded moments happen live, so authorship is never in question. Second, teaching critique of machine-generated readings — handing students an AI summary of a novel and making them find where it is thin is a literature lesson in itself.
Neighboring jobs face a similar mix. Compare the narrative and task split for foreign language and literature teachers, communications teachers and history teachers, all in the postsecondary teachers family and the wider education sector. On our headline measure, this job scores 63 out of 100 (higher is safer). To see how that sits against another role you are weighing, put the two side by side in our comparison tool, or browse the jobs that most need a person.