Skip to content
STRATUSWORKFORCE SCAN

Physical Therapist Assistants

In short: today's best AI models could do about 12% of this job's working time (projected from METR's April 2026 measurement and Epoch's November 2026 index) if the work were set up for them, and about the same by the end of 2028 if progress keeps its long-run pace. The conversations, in-person and hands-on work, about 88% of the time, stays with people.

Compare with another job

Assist physical therapists in providing physical therapy treatments and procedures. May, in accordance with state laws, assist in the development of treatment plans, carry out routine functions, document the progress of treatment, and modify specific treatments in accordance with patient status and within the scope of treatment plans established by a physical therapist. Generally requires formal training.

112,430 people do this job in the US. $68,380 a year.

Reviewed on September 30, 2026: what each task is made of, and what handing it over would ask.

A typical 40-hour week

  • AI could take on0 h
  • A person checks AI's work5 h
  • Desk work AI can't do yet0 h
  • Stays with people35 h

Physical Therapist Assistants, now: of a 40-hour week, AI could take on 0 hours, 5 hours become checking its work, 0 hours are desk work it can't do yet, and 35 hours stay with people.

The largest pieces: Document patient progress (about 2 h, a person checks); Observe patients and report responses (about 2 h, a person checks); Confer on patient treatment (under an hour, a person checks).

Estimates for a typical week of this job, if progress keeps its long-run pace, counting work the best models finish four times in five. Every task is below.

now
12%

of the working time. Now is projected from METR's April 2026 measurement and Epoch's November 2026 index.

What the work is made of
  • 1%
  • 11%
  • 2%
  • 29%
  • 58%
0%

of the work: Anthropic's measure of what people already bring to Claude at work (2025). It counts help with part of a task, so it can be higher than the share within reach.

Half the desk work within reach
Already

On the chosen pace and reliability. Estimate.

Share of the working time within reach, now to 2030

0%25%50%75%100%2027202820292030NowDesk work: 12% of the time
Chosen to Ceiling: the share
DateNow
NowEnd of 2030
Change assumptionsLong-run pace, four in five

What each choice means: , , , , . Every pace starts from .

At this date and pace, models finish tasks up to 2.4 work days long four times in five.

What handing it over would take, now

The 12% within reach splits by what a model would need. The dates, pace and reliability follow the controls above. Estimates.

1%

of the working time. The material can be handed to a model and the result handed back.

11%

of the working time. The model has to be connected to the system the work lives in.

, not counted
2%

is conversation short enough by length. This site leaves conversation with people, so it is not in the total.

How far this has come

With GPT-4, measured in March 2023, 0% of this job's working time was within reach at . With Claude 3.7 Sonnet, measured in February 2025, 3%. With Claude Mythos Preview (early), measured in April 2026, 12%. Today's estimate: 12%.

0%5%10%15%20%2023202420252026GPT-4, March 2023: 0% within reachGPT-4 Turbo (Nov 2023), November 2023: 0% within reachGPT-4o, May 2024: 0% within reachClaude 3.5 Sonnet (June 2024), June 2024: 0% within reacho1-preview, September 2024: 0% within reachClaude 3.5 Sonnet (Oct 2024), October 2024: 0% within reacho1, December 2024: 1% within reachClaude 3.7 Sonnet, February 2025: 3% within reacho3, April 2025: 9% within reachGPT-5, August 2025: 10% within reachGemini 3 Pro, November 2025: 11% within reachClaude Opus 4.5, November 2025: 11% within reachGPT-5.2, December 2025: 12% within reachClaude Opus 4.6, February 2026: 12% within reachClaude Mythos Preview (early), April 2026: 12% within reachToday (estimate), November 2026: 12% within reach
Each step is a new best model measured by METR, four in five; the dashed end is today's estimate. Point at a dot for the model.

A backcast: today's task data with the best model METR had measured by each date. It shows how fast the models have moved, not what anyone forecast at the time.

Measured use over time

Of the Claude.ai conversations Anthropic sampled and tied to a work task, 0.011% went to this job's tasks in May 2026, the 397th largest of the 447 jobs that showed up at all; 1 of its 21 tasks showed up. Within each of Anthropic's methods, never across a change: 0.00096% to 0.00081% (August 2025 to February 2026); 0.011% to 0.011% (April to May 2026).

0%0.01%0.02%20252026released February 2025: nonereleased March 2025: nonethe week of Aug 4, 2025: 0.00096%the week of Nov 13, 2025: 0.0018%the week of Feb 5, 2026: 0.00081%April 2026: 0.011%May 2026: 0.011%
Each dot is one release's sample. A line joins releases made by the same method; a change of method starts a new line. Point at a dot for its window.
Each release
ConversationsShareRankTasks seen
Released February 2025nonen/a0 of 21
Released March 2025nonen/a0 of 21
The week of Aug 4, 20250.00096%654 of 6551 of 21
The week of Nov 13, 20250.0018%657 of 6721 of 21
The week of Feb 5, 20260.00081%677 of 6771 of 21
April 20260.011%381 of 4341 of 21
May 20260.011%397 of 4471 of 21

Measured, not estimated: Anthropic's samples of Claude.ai conversations, each tied by a model to an O*NET task, from the Anthropic Economic Index. It says where Claude's use goes, not how much of this job's working time it covers, so it is not on the same scale as the reach above. The releases differ in their sample, their classifier and the smallest share they report (the June 2026 file rounds each task to a hundredth of a percent, so a smaller task reads as none), which is why a change of method starts a new line. At least 96% of each older release's task-tied use matched this site's tasks, and all of the newest. Claude.ai only, so use through other tools and other companies' models is not in it. Every job ranked.

Robots and the physical work

Anthropic rates 68% of this job's working time as physical work. For 50% of the working time, a robot has been shown doing most of the task somewhere: 50% in a purpose-built cell. 0.50 of 3, the 313th highest of the 923 jobs.

  • Not physical 32%
  • No robot shown 18%
  • In a purpose-built cell 50%

Anthropic's rating, not this site's estimate: Claude Opus 5, searching the web, rated each physical O*NET task by the kind of setting in which a robot it could cite (deployed, sold or demonstrated) does most of the task, weighted by its own estimate of each task's share of the time (What work can robots do?, September 30, 2026, data CC BY 4.0; every job ranked). A robot shown somewhere is not a robot in your building: cost, setup and safety decide that. The share within reach on this page still counts no body work. The robot behind each rated task is under the task below.

What changes for the people in this job

From the task-by-task review, by the end of 2028 on the long-run pace at four in five. It says which parts of the work stay with people and which change, not whether anyone keeps a job: that depends on decisions no data set can see.

What stays with people

88% of the working time: the conversation, in-person and hands-on parts of the tasks.

  • Guide patients through exercises, the in-person work, 18%
  • Teach body mechanics and mobility, the in-person work, 16%
  • Administer therapy and modalities, the hands-on work, 12%
  • Observe patients and report responses, the hands-on work, 10%
  • Clean and store equipment, the hands-on work, 4%
  • Instruct caregivers and families, the in-person work, 4%

What becomes checking the machine

11% of the working time: desk work a model could do by then, in tasks where a mistake costs something or a licensed person signs. A person checks it.

  • Document patient progress, severe if wrong, 5%
  • Observe patients and report responses, a licensed person signs, 4%
  • Confer on patient treatment, severe if wrong, 1%
  • Monitor equipment and record treatment, severe if wrong, 0.8%

What to learn

  • Judgment where errors are severe. Some tasks are severe if wrong. Knowing when not to trust an output matters more there than speed.
  • Signing off under your license. Some tasks need a licensed, certified or sworn person to do or sign them. The license, and the judgment it stands for, stays with the person.
  • The hands-on craft. About 58% of the time is hands-on work, which no model here does; robots are rated separately on this page.
  • The conversations. About 30% of the time is conversation and in-person work, which this site leaves with people: listening, explaining, persuading, calming.
  • Dealing with people outside the organization. Customers, patients, suppliers or the public: trust with them is built by a person.

Your own version of this job: read it against your job description. The headline is free; the full personal report, every task with its timeline and what to learn, is $19 (a sample).

The tasks

Sorted by . 21 tasks from .

Task
Instruct, motivate, safeguard, and assist patients as they practice exercises or functional activities.Coaching and safeguarding the patient through the exercises.Done hourly or more by most who do it. Robots: in a purpose-built cell, Lokomat / LokomatPro (Hocoma) (sold). Anthropic's rating.18%25 min13 min to 50 minIn person 60%, hands-on 40%Stays with people
Instruct patients in proper body mechanics and in ways to improve functional mobility, such as aquatic exercise.Instruction in person, with hands-on help.Done hourly or more by most who do it. Robots: no robot shown. Anthropic's rating.16%25 min13 min to 38 minIn person 70%, hands-on 30%Stays with people
Observe patients during treatments to compile and evaluate data on their responses and progress and provide results to physical therapist in person or through progress notes.Watching the patient in treatment, with the notes for the therapist.Done hourly or more by most who do it. 14%21 min8 min to 38 minHands-on 70%, systems 30%30%Desk part half by now
Administer active or passive manual therapeutic exercises, therapeutic massage, aquatic physical therapy, or heat, light, sound, or electrical modality treatments, such as ultrasound.Manual exercises, massage and modality treatments by hand.Done hourly or more by most who do it. Robots: in a purpose-built cell, Harmony SHR (Harmonic Bionics) (sold). Anthropic's rating.12%25 min13 min to 50 minHands-on 100%Stays with people
Document patient information, such as notes on their progress.Notes entered in the record.Done hourly or more by most who do it. 5%8 min4 min to 21 minSystems 100%100%Half by now
Clean work area and check and store equipment after treatment.Tidying after treatment.Done hourly or more by most who do it. Robots: in a purpose-built cell, Peanut cleaning robot (sold). Anthropic's rating.4%8 min4 min to 17 minHands-on 100%Stays with people
Communicate with or instruct caregivers or family members on patient therapeutic activities or treatment plans.Instruction in person and by phone.Done daily by most who do it. 4%17 min8 min to 33 minIn person 70%, conversation 30%Stays with people
Secure patients into or onto therapy equipment.Strapping the patient into the equipment.Done hourly or more by most who do it. Robots: no robot shown. Anthropic's rating.3%7 min3 min to 13 minHands-on 100%Stays with people
Confer with physical therapy staff or others to discuss and evaluate patient information for planning, modifying, or coordinating treatment.Conversation with the therapy staff in the clinic, around the chart.Done daily by most who do it. 3%17 min8 min to 25 minIn person 60%, documents 40%40%Desk part half by now
Prepare treatment areas and electrotherapy equipment for use by physiotherapists.Areas and electrotherapy equipment readied.Done hourly or more by most who do it.Robots: in a purpose-built cell, Figure 03 humanoids with Helix-02 policy (demonstrated). Anthropic's rating.3%13 min4 min to 25 minHands-on 100%Stays with people
Train patients in the use of orthopedic braces, prostheses, or supportive devices.Teaching the device in person, hands on.Done more than yearly by most who do it. Robots: no robot shown. Anthropic's rating.3%21 min13 min to 33 minIn person 60%, hands-on 40%Stays with people
Transport patients to and from treatment areas, lifting and transferring them according to positioning requirements.Lifting and moving patients by hand.Done several times a day by most who do it. Robots: no robot shown. Anthropic's rating.3%10 min4 min to 21 minHands-on 100%Stays with people

Within reach, per task: the share of the task's time whose instances are short enough for a model to finish at the chosen date, pace and reliability. Half by: the date half of that time comes within reach. Made of: how the task's time splits among five kinds of doing. Only the documents and systems parts can count as within reach, so a task that is 30% documents can be at most 30% within reach.

Jobs that share skills with this one

From O*NET's related occupations. Within reach now and by the end of 2028, long-run pace, four in five.

Where this job works

The industry groups that employ the most of it (BLS, May 2025).

Where the most of these jobs are

The metro areas with the most people in it (BLS, May 2025).

Questions about this job

How much of the work of Physical Therapist Assistants is within reach of AI now?
An estimated 12% of the working time is within reach of AI models now and 12% by the end of 2028, on METR's long-run pace at four in five. Within reach means a model could finish work that long at that reliability. It is not a forecast of jobs lost or of savings.
Which tasks of Physical Therapist Assistants come within reach first?
Weighted by the time each takes, the tasks with the most within reach now (projected from METR's April 2026 measurement and Epoch's November 2026 index) are: Document patient progress (100% of the task's time now); Observe patients and report responses (30% of the task's time now); and Confer on patient treatment (40% of the task's time now).
Which tasks of Physical Therapist Assistants stay with people?
Tasks that are people work or body work are not counted as within reach at any date. The largest: Guide patients through exercises; Teach body mechanics and mobility; and Administer therapy and modalities. Together such tasks are about 74% of the working time, and the people and body parts of the other tasks stay with people too.
What should Physical Therapist Assistants learn as AI takes on more of the work?
By the job's own task-by-task review, the first things to learn are judgment where errors are severe; signing off under your license; and the hands-on craft. By the end of 2028 on the long-run pace at four in five, about 88% of the working time stays with people, 11% becomes checking a model's work and 1% could be handed over with lighter checks. It says which parts of the work change, never whether anyone keeps a job.
Do people already use AI for the work of Physical Therapist Assistants?
In the Claude.ai conversations Anthropic sampled for May 2026 and tied to a work task, 0.011% went to this job's tasks, the 397th largest share of the 447 jobs that showed up; 1 of its 21 tasks appeared. Anthropic's labor market measure puts the share of the job's tasks seen in real Claude use at work at 0%. These are measures of use, not of how much of the work AI could finish, and they cover Claude only.
Can robots do the physical work of Physical Therapist Assistants?
Anthropic's robot exposure rating (September 2026) finds 68% of the working time is physical work, and that a robot has been shown doing most of the task, somewhere, for 50% of the working time. Its index is 0.50 of 3. One example: Lokomat / LokomatPro for "guide patients through exercises". A robot shown somewhere is not one in every workplace; this site's own estimates still count no physical work as within reach.
How is this estimated?
From the 21 tasks O*NET lists for the occupation, each with an estimated length checked against GDPval and a share of the working time, reviewed one task at a time on September 30, 2026 for what each is made of (documents, business systems, conversation, in person, hands-on), and METR's measurements of how long a task the best AI models can finish, carried forward to the date from the latest, Claude Mythos Preview (early) in April 2026. Data checked November 10, 2026.

How the estimates are made

This job in your state or metro area
Embed these figures on your site

A small card with the share of working time within reach now and at the end of 2027, 2028 and 2030, the date and a link back. Free under CC BY 4.0; the card carries the credit.

See the card

Your own version of this job

The page above is the occupation. Your job is your title and what you actually do: give both, and they are read against these tasks in your browser, nothing sent anywhere. The headline is free: how much of your working time is within reach now and by the end of 2028, and how many of the tasks your description names.

The full personal report is $19, to print or save as a PDF: every task with when it comes within reach, which ones your description names, the parts of your description the task list does not cover, what stays with you, what becomes checking the machine and what to learn.