Why hands-on rehab keeps a person in the room
A physical therapist assistant spends most of the day within arm’s reach of a patient. You walk someone through gait training with a walker, feel how much weight they shift onto a bad hip, and change the drill mid-set when the form slips. You apply modalities too: ultrasound, electrical stimulation, hot and cold packs, manual stretching. Software can prompt the next exercise. It cannot take the weight or read a wince before the patient speaks.
The second reason is the supervision loop. A PTA carries out a plan of care written by the physical therapist, then reports what actually happened in the session: the knee that buckled, the shoulder that moved further than last week, the patient who hid pain because they want to go home. That reporting runs on touch, observation and conversation, and a licensed clinician signs off on it.
Demand matters as well. The Bureau of Labor Statistics counts about 112,430 physical therapist assistants in the United States, with median pay of $68,380, and projects employment growth of 23% from 2025 to 2035 (BLS, 2025). An older population needs more rehab after strokes, joint replacements and falls. That is a lot of sessions that someone has to run in person.
What AI handles, what it assists, what stays with the assistant
Start with the work software can do on its own. Our task split puts 0% of task time in that group: mostly the paperwork around care, such as writing up treatment notes and progress records, and handling scheduling and billing details between visits. Speech-to-text and note drafting tools already shorten that part of the day. The share of task time AI can handle today is the Can AI do it? score of 10 out of 100, and how coverage is measured explains what counts.
Next, the assisted group, at 15% of task time. Home exercise programs are a clear case: an app can send the right drills, count reps from a phone camera and flag missed days. Progress tracking is another, where sensors and motion analysis give the therapist numbers on range of motion or walking speed that used to be eyeballed.
The rest, 85% of task time, stays with the assistant. Manual therapy and hands-on modalities sit here, and so does training a patient to use crutches, a cane or a prosthesis safely on stairs. So does the part nobody automates well: keeping a scared or discouraged patient working through the session.
What the evidence actually tests
There is no direct head-to-head test of AI against physical therapist assistants doing this job. That is why the Is it better than a person? question carries an evidence grade of D and no parity number at all. Grading rules are set out in how parity is graded, and a grade at that level means the comparison has not been measured, not that AI performed badly.
What would settle it is specific. A trial would have to compare patients treated by AI-guided or robot-assisted sessions with patients treated by a PTA, on outcomes clinicians already track: gait speed, range of motion, adherence to the home program, re-injury rates, discharge timing. Rehab robots exist in research settings and some clinics, but they are used under a clinician’s hands, not instead of one. Until outcome data arrives, the honest position is that the comparison is untested here.
When the picture could change
Most likely after 2045 (8 in 10 of our scenarios). See how the replacement year is estimated for what that window is built from.
Two things could pull it earlier. The first is robotics: about two thirds of this job is physical work, and the robotics panel on this page places it in the dexterous humanoid tier, so a real jump in machine handling and balance would matter more here than better chatbots. The second is cost pressure. Clinic software subscriptions are cheap next to staffing a session, so anything that lets one therapist supervise more patients at once will be tried first.
Two things hold it back. Licensure and supervision rules tie treatment to a credentialed person who is accountable for what happens to the patient. And the physical risk is real: a machine that mishandles a post-surgical shoulder or lets a stroke patient fall is a liability problem, not a software bug. Insurers and state boards move slowly on both.
Good to know: the pressure in rehab shows up as fewer easy entry-level hours, not as clinics running without assistants.
How to stay needed as a PTA
Lean into the tasks that stay with people. Gait and transfer training with assistive devices is one. Manual therapy and modality work where you adjust by feel is another. The third is the judgment call you make mid-session, when the planned progression is wrong for the patient in front of you, and the write-up you give the supervising therapist afterward.
Two skills are worth adding. Learn to use motion tracking and home-exercise apps well enough to set them up for patients who are not comfortable with technology; that makes you the person who makes the tools work. And get sharper at documentation review, so you can catch what an AI note draft got wrong before it goes in the chart.
If you want to see where nearby roles land, compare this page with physical therapist aides, occupational therapy assistants and physical therapists, or put any two side by side with the job comparison tool. The wider therapy assistants and aides family and the healthcare sector page show the same scores across related roles, and the list of jobs that most need a person puts rehab work in context. Every figure on this page is built the same way, as set out in the scoring method.