Why law teaching keeps a person at the front of the room
Will AI replace law teachers? The task mix says no, though parts of the week are already moving. Law teaching runs on live argument. A professor initiates and moderates classroom discussion, pushes a student to defend a position, then changes the hypothetical when the answer gets too easy. That exchange is the point of the method, and it depends on reading the room in real time.
The second anchor is judgment about people. Advising students on course choices and career paths, supervising clinical and field work, and sitting on faculty and admissions committees all carry responsibility a tool cannot hold. Accreditation, grading appeals and bar outcomes attach to a named instructor.
What is moving is preparation and paperwork. Drafting lecture outlines, summarizing new case law, building practice questions and giving first-pass feedback on legal writing are all jobs software now takes a decent swing at. That is task erosion, not a vanishing occupation. Our Still needs a human score for this job is 64 out of 100 (higher is safer), and you can see how that figure is built on the methodology page.
What AI does, what it assists with, and what stays with faculty
Some work AI can already take on its own. Compiling reading lists from current literature, producing draft exam items and multiple-choice banks, and keeping attendance and grade records need little supervision once a template exists. Share of task time in that group: 9%.
A larger block is assisted work, where the tool drafts and a person decides. Preparing lectures and course materials, and marking student papers and essays, both fall here: a model can summarize a 60-page opinion or flag a weak argument in a brief, but a professor sets the standard, catches hallucinated citations and owns the grade. Share of task time in that group: 52%. The coverage figure of 42 measures how much task time AI can handle today; the coverage method explains what counts.
Then there is the work left to people: moderating Socratic discussion, supervising clinics and moot court, mentoring students through academic trouble, and service on committees and in faculty governance. Share of task time: 39%. None of it needs a robot, which is why hardware is not the limit here. The limits are judgment, accountability and trust.
What the evidence actually shows
No study has yet tested an AI system against law faculty at the job itself. Our evidence grade for quality parity is D, and a D grade means not measured, so we publish no parity number for this occupation. Claims that models pass bar-style exams say something about answering questions, not about running a semester.
What would settle it is specific: a comparison of sections taught by faculty against sections run mainly by AI, scored on blind-marked student writing, exam performance and later bar results; and a marking study where graders cannot tell human feedback from machine feedback. Until that exists, treat confident predictions in either direction as opinion. The quality parity method sets out the bar we need before a number goes up.
Market data gives some context. Our dataset puts US employment for this occupation at about 20,060 with median pay of $128,500 and projected growth of 2.6% through 2035 (BLS). Small, specialized, credential-gated fields rarely shed roles quickly, but hiring at the entry rungs is where pressure usually lands first.
When the picture could change
Most likely between 2034 and 2044 (8 in 10 of our scenarios). The replacement-year method explains what that window is measuring and how we build it.
Two things could pull it earlier. Cost is one: licensing software for course prep and first-pass grading is far cheaper than adding a faculty line, and budget-pressed schools notice. Adoption is the other: once legal research assistants and writing-feedback tools are standard in practice, schools have a reason to hand routine instruction to them and spend faculty time on supervision.
Two things hold it back. Accreditation and institutional rules require a responsible instructor of record, with named accountability for grades and for what students are taught. And reliability still bites, because a fabricated citation or a confidently wrong reading of a statute is a serious failure in this field. Trust has to be earned case by case, not announced.
Good to know: the hardest part of this job to automate is not knowing the law, it is deciding what a particular student needs to hear next.
How law faculty stay needed
Lean into the tasks the task split leaves with people. Run discussion that cannot be scripted, with hypotheticals built from current disputes. Take on clinical and moot court supervision, where students are judged on live performance. And take governance seriously: curriculum design, admissions and standards work is where a department’s direction is set.
Two skills matter beyond that. First, supervising AI output: knowing how to check a generated case summary, spot a bad citation and teach students to do the same. Second, assessment design that resists copy-paste answers, which usually means oral components, in-class work and layered drafts.
Nearby teaching jobs face a similar mix, and comparing them is useful. Look at Criminal Justice and Law Enforcement Teachers, Political Science Teachers and Philosophy and Religion Teachers. The wider postsecondary teachers family page shows the pattern across disciplines, and the education sector page covers the institutions that employ them.
Next step: put two of these jobs side by side with the compare tool, or see where teaching roles land in our list of jobs most likely to still need a human.