← Risk register SOC 29-1123 · reviewed 2026-08-11

Physical Therapists

267,330 US workers · median $102,760/yr · Healthcare

SAFE

The core of the job is hands-on: palpating tissue, guiding a post-surgical knee through range of motion, spotting a stroke patient's gait deviation in real time, and adjusting load mid-session based on how the body responds. AI can draft the SOAP note, pick exercise templates, flag insurance-authorization language, and triage outcome data — all real time savings, none of it the treatment. State licensure plus personal accountability for patient safety during mobilization means a licensed human signs the plan of care and takes the fall if a patient falls.

10-year outlook: Demand grows with an aging population while AI absorbs documentation and exercise-prescription drudgery; the main pressure comes from payers pushing volume onto assistants and app-based home programs, not from AI replacing the clinician.

US employment, 2019–2025+14.6%
233,350267,330 workers

Dipped in 2020, then grew past where it started.

Median pay $89,440 → $102,760 -8.1% in real terms (nominal +14.9%, less ~25% US inflation over the period)

The job count is not the verdict

This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.

So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.

BLS projection, 2024–2034

+10.9% 267,200 → 296,400 on the projections basis

Hard to automate, and growing

The work resists current AI and the BLS projects +10.9% more of these jobs by 2034. Note that safe does not mean well paid — several of the fastest-growing resistant occupations are among the lowest paid on the register.

Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.

~13,200 openings a year on average, including replacing people who leave.

One email if this score changes. Watch as many occupations as you like from the same address — no account, and nothing is sent on a schedule, only when a verdict actually moves.

Also known as — 23 job titles this covers

Titles reported by people doing this work, from the US Department of Labor's O*NET survey. If your job title is here, this page is about your work even though the name doesn't match.

TherapistPhysiotherapistKinesiotherapistSchool Physical TherapistDoctor of Physical Therapy (DPT)LPT (Licensed Physical Therapist)Acute Physical Therapist (Acute PT)Registered Physical Therapist (RPT)Sports Physical Therapist (Sports PT)Travel Physical Therapist (Travel PT)Geriatric Physical Therapist (Geriatric PT)Home Care Physical Therapist (Home Care PT)Inpatient Physical Therapist (Inpatient PT)Pediatric Physical Therapist (Pediatric PT)Pulmonary Physical Therapist (Pulmonary PT)Acute Care PT (Acute Care Physical Therapist)Orthopedic Physical Therapist (Orthopedic PT)Outpatient Physical Therapist (Outpatient PT)Home Health Physical Therapist (Home Health PT)Skilled Nursing Facility Physical Therapist (SNF PT)Cardiopulmonary Physical Therapist (Cardiopulmonary PT)Outpatient Travel Physical Therapist (Outpatient Travel PT)Outpatient Orthopedics Physical Therapist (Outpatient Ortho PT)

Score — 86/100 resistance

Holding it up: embodiment (19/20). Weakest point: task resistance (15/20).

Five dimensions, 0–20 each, summed. Higher means more protected. The arithmetic is shown so you can check it: 15 + 19 + 17 + 17 + 18 = 86. · Scored 2026-08-11, and re-examined when evidence accumulates rather than on a schedule.

Task resistance 15/20

Tasks largely resist digitisation Manual therapy, joint mobilization, and gait retraining can't be delegated to software — the parts that can (documentation, HEP handouts, initial outcome-measure scoring, prior-auth letters) are maybe a fifth of the workday, which is why this sits at 15 rather than 18: the eval interview, goal-setting, and progress-note reasoning are increasingly template-and-LLM assisted even though the treatment itself isn't.

Embodiment 19/20

Hands-on in uncontrolled environments You are physically under a 200-pound patient during a sit-to-stand transfer, feeling for end-feel at a stiff shoulder capsule, and catching someone whose knee buckles on a ramp — 19 rather than 20 only because a slice of practice (telehealth follow-ups, wound-care consult review, some pediatric parent coaching) happens without your hands on anyone.

Liability shield 17/20

Licensed human required and personally liable Every state licenses PTs under a practice act, you carry your own NPI and malpractice policy, and in direct-access states there is no physician co-signature between you and a missed red flag — the plan of care has your license number on it, which is why this is 17 and not the 12 of a role where a supervising MD absorbs the exposure.

Trust premium 17/20

The human relationship is the product A 12-visit episode for a rotator cuff repair means the patient sees you two or three times a week for six weeks and tells you about their pain, their fear of re-tearing, and whether they actually did the exercises — adherence is the treatment effect, and it comes from that relationship, held just short of 20 because staffing models rotate patients across PTs and techs and patients still come back.

Judgment & accountability 18/20

Exists to be accountable for ambiguous calls You decide when a post-op knee is ready to progress load, when a new-onset unilateral calf pain means stop and send to the ED for DVT, and when to discharge against a payer's remaining visits — those are consequential calls made in the room with incomplete information, which is 18 rather than 14 because a wrong one causes immediate physical harm.

Confidence: high · reviewed 2026-08-11 · how scoring works

What this job involves — and which parts are yours

The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.

AI already does these at usable quality

These still need a person

Active moats on the surviving side: embodiment, licensure, trust, judgment

How to future-proof this job

Where to go deeper on what this job runs on: Khan Academy — reading and vocabulary, all levels, free free · Coursera — active listening and communication skills free to audit · Toastmasters — public speaking practice at local clubs worldwide low · Coursera — critical thinking and logic, audit free free to audit · Coursera — communication and interpersonal skills free to audit · Coursera — customer service and client-facing skill courses free to audit

All 35 skills ranked by how many jobs they open →

What would move this back up — beyond any one person

The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 93/100, still SAFE.

4 specific changes that would raise this score
  • already happening task resistance +2

    Documentation, prior-auth narrative, HEP handout generation and outcome-measure scoring are the routine tier and are being absorbed first (Netsmart/WebPT ambient scribe products). If that tier is stripped out, the residual day is palpation, manual therapy, real-time load titration and gait retraining — the part with no software substitute. Score rises because the denominator shrinks, not because AI got worse.

  • already happening trust premium +2

    Growth of cash-pay/out-of-network manual therapy and performance clinics where the buyer is explicitly paying for one-on-one hands-on time with a named clinician rather than insurance-reimbursed group treatment — visible in the cash-based PT segment and in the backlash against high-volume 3-patients-per-hour clinics. Also observable if payers begin covering app-delivered exercise (Hinge Health, Sword) as the default and hands-on care becomes the premium tier patients top up for.

  • already happening judgment accountability +1

    If direct-access referral authority keeps expanding (all 50 states now have some form; the fight is over imaging referral and unrestricted evaluation), the PT becomes the first-contact differential diagnostician who must rule out red flags — cauda equina, DVT, fracture, cardiac referral pain — without a physician gate. That shifts consequential ambiguity onto the PT rather than the referring MD.

  • plausible liability shield +2

    Medicare's therapy supervision rules tightening rather than loosening — e.g. CMS reversing the 2022 PTA general-supervision allowance or the de minimis modifier policy, so that billable therapy minutes require direct on-site licensed PT presence rather than assistant delivery with remote oversight. Equally, state practice acts adding explicit language that AI-generated exercise prescriptions or telehealth home-program plans require a named licensed PT of record who retains personal liability for adverse events (falls, post-op dehiscence).

The limit. At 86 the headroom is arithmetic more than substantive — embodiment at 19 and judgment at 18 are effectively maxed. The realistic risk here is not displacement of the licensed PT but volume compression: digital MSK programs substituting for the low-acuity caseload that currently supports clinic economics, shrinking the number of PT jobs while leaving each remaining job more resistant. A rising resistance score and a falling headcount are compatible.

These are conditions, not forecasts — what would have to happen, not what will. Specific rules, cases and bills are named so you can go and check whether they exist and where they stand; verify before relying on any of them. Nothing here is legal or financial advice.

Where this work is, and what it pays there

BLS metro figures for 388 areas. The verdict above does not change by city — the rubric judges what the work involves, not where it happens — but pay and headcount do, and the national median hides a very wide range.

Most of these jobs

New York-Newark-Jersey City, NY-NJ 21,240 $107,650 +5%
Los Angeles-Long Beach-Anaheim, CA 9,750 $108,170 +5%
Chicago-Naperville-Elgin, IL-IN 7,880 $105,720 +3%
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD 6,250 $105,790 +3%
Dallas-Fort Worth-Arlington, TX 6,100 $107,780 +5%
Boston-Cambridge-Newton, MA-NH 5,330 $103,550 +1%
Miami-Fort Lauderdale-West Palm Beach, FL 4,950 $100,870 -2%
Houston-Pasadena-The Woodlands, TX 4,680 $107,700 +5%

Best paid

San Francisco-Oakland-Fremont, CA 3,140 $141,980 +38%
Vallejo, CA 300 $141,530 +38%
Santa Rosa-Petaluma, CA 290 $140,100 +36%

Percentages are against this occupation's national median of $102,760. Counts are jobs in that metro, not vacancies. Metros where the BLS suppressed the cell are absent rather than shown as zero.

Who is actually doing this — nobody, on the record

We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.

That is worth saying out loud next to a score of 86. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.

Read that as a gap in the reporting we can see, not proof of absence — the dispatch runs on English-language feeds and misses plenty. If you know of a case, tell us, or add a field report from inside the job.

Quick take — do you do this job?

Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.

Self-reported and unverified — a sentiment signal, not a survey. One response per person per occupation; you can change your answer.

Field reports — what people say has changed

No field reports yet. A written account takes a paragraph rather than a tap, goes to an editor before it appears, and is the one thing on this page the rubric cannot produce on its own.

File a field report

Concrete beats general: a tool that arrived, a task that moved, a headcount decision you watched happen. Don't include anything that identifies you or your employer if that would put you at risk.

Watch this verdict
Kept current

Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.