Why the work stays in the room
Choreography happens on specific bodies, in a specific room, under a deadline. A choreographer designs dances for a company, a musical, an opera or a film, then directs rehearsals until the cast can actually perform them. That second part is the job. Steps on a screen are a suggestion; steps that a particular dancer can hit eight shows a week are a result.
The adjustment work is constant. One dancer has a tight hamstring. Another can jump higher than the phrase was built for. The director changes the scene length the night before tech. Auditioning performers and advising dancers on how to move safely both depend on watching a person in front of you and changing your mind in seconds. Software can propose movement. It cannot carry the responsibility for a lift that goes wrong.
That is why the question will ai replace choreographers reads differently than it does for desk jobs. The exposure here is narrow and specific: screen and short-form work, reference material, and the documentation around a piece. The live core is slower to move.
What AI does, what it assists, what it leaves to people
A small slice of the task time is work AI can take on alone: 0% of the exposed time. That is mostly desk work around the dance, not the dance. Reading and summarizing story lines or musical scores to find structure is one example. Writing up movement sequences and keeping records of a piece is another.
A similar slice is assisted work, at 10% of the exposed time: a tool speeds up a task but a choreographer still decides. Choosing and trimming music or sound to accompany a dance is faster with audio tools. Reviewing rehearsal footage, tagging takes and comparing versions is faster with motion capture and video analysis. Generative motion models can also throw out variations to react against, which is closer to a sketchpad than a collaborator.
Everything else sits with the person: 90% of task time. Directing rehearsals and teaching dancers the steps is first. Auditioning and casting is second, because it is a judgment about a human being, not a filter. Coaching dancers on alignment, timing and safe technique is third. The overall coverage figure, 12 out of 100, is built from that split; how coverage is measured explains the method behind it.
Good to know: the robotics panel above puts this job in the lowest hardware tier, so nothing here waits on a machine that can move like a dancer.
What the evidence actually shows
The evidence grade for this job is D. In plain terms, no study has yet tested AI-made choreography against a working choreographer’s under fair conditions, so this page gives no parity number. Claims in either direction, from tool makers or from critics, are opinion until somebody runs the test.
A fair test is easy to describe. Take a brief a company would really issue, have both a choreographer and a generative system produce phrases, stage both on the same dancers with the same rehearsal hours, then have artistic directors and audiences judge the results blind. Measure rehearsal time, injury reports and how well the cast retains the material. Until something like that exists, the honest position is that we don’t know how close the output is. You can read how we grade evidence on the quality parity page, and the full scoring method covers the rest.
When the picture could change
Most likely between 2037 and 2056 (8 in 10 of our scenarios). What that window measures is set out on the replacement year page.
Two things could pull it earlier. Screen work is the soft edge: filmed, animated and game movement can be built from motion capture libraries and generative motion models, where no live cast is involved and no one has to repeat it nightly. Cost is the other. The cost panel above shows the tooling side running from tens to a few thousand dollars, against a far larger figure for the human equivalent, which makes experimentation cheap for producers working at volume.
Two things hold it back. Live staging stays tied to a named cast and a safety duty that nobody has worked out how to assign to software. And the market is small: BLS counts about 2,860 choreographers in the US and projects 3.1% employment growth from 2025 to 2035, with median pay of $55,310 (BLS, 2025). A small, slow market gives tool builders little reason to chase the live end of the job. The real pressure shows up earlier in a career, where the paid screen and commercial gigs that used to fund new choreographers get cheaper to fake.
How to stay needed
Lean into the tasks that only work with people present. Run the room: rehearsal direction, note-giving and fixing a phrase live is the part nobody can outsource. Own casting, because auditioning is a judgment about potential, not a match score. Take the coaching seriously, including safe technique and injury prevention, since that is where trust with a company is built.
Two skills raise your floor. First, restaging and notation, so your work can travel to other companies and other casts under your name. Second, working fluently with video, motion capture and editing tools, so you lead the screen version of a project instead of being briefed out of it. Credit and consent terms for captured movement are worth reading closely before you sign.
If you are weighing options, three nearby jobs are worth a look: dancers, music directors and composers and producers and directors. You can put any two of them next to each other with the job comparison tool, see the wider group on the performers and entertainers family page, or read the whole industry picture in arts and entertainment. For context on where this job sits against everything else, the jobs that most need a person list is the place to start.