EXPOSED
A large share of a physicist's week — literature review, simulation and analysis code, fitting models to data, drafting papers and grant text — is exactly what current AI does at usable quality, and computational/theory work is where that bite is deepest. What resists is designing an experiment that has never been run, building and debugging real apparatus (cryostats, beamlines, optics benches, detectors), and deciding which anomaly is physics versus instrument artifact. Modal employment is national-lab, federal, and industrial R&D; medical physicists inside this bucket carry board certification and radiation-safety accountability that most other physicists do not.
Dipped in 2020, then grew past where it started.
Median pay $122,850 → $172,250 +12.2% in real terms
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
+4%
Percentage only. The projection counts a different population from the 20,430 above — it includes self-employed workers, which for this occupation is most of them, so the two headcounts are not comparable.
Growing, and only partly exposed
The BLS expects +4% more of these jobs by 2034, and at 46/100 the work is only partly exposed — some tasks are automatable, the core of the job is not. Nothing here is in tension.
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.
~1,700 openings a year on average, including replacing people who leave.
PhysicistScientistRheologistAerophysicistAerodynamicistAstrophysicistCloud PhysicistSpace PhysicistFluid DynamicistHealth PhysicistPlasma PhysicistRocket ScientistThermodynamicistElectrodynamicistMedical PhysicistNuclear PhysicistNuclear ScientistOptical ScientistQuantum PhysicistResearch PhysicistResearch ScientistMass SpectroscopistMolecular PhysicistRadiation Physicist
Holding it up: judgment & accountability . Weakest point: liability shield .
Mixed — a routine tier and a judgment tier An 11 reflects the split inside the actual job: the analysis pipeline, Monte Carlo work, Mathematica derivations, and manuscript drafting that fill most of a computational or theory physicist's week are already being done at usable quality by models, while aligning an optics bench, chasing a vacuum leak, commissioning a beamline shift, or judging that a 3-sigma bump is a bad calibration run has no digital substitute — enough irreplaceable work to keep it out of the single digits, not enough to reach the mid-teens where wet-lab and field scientists sit.
Some physical or field component A 9 puts physicists above pure desk work because cryostat assembly, detector installation, cleanroom fabrication, and accelerator hall access are real hands-on hours, but the environments are controlled — a temperature-stabilised lab or a shielded vault, not a rig floor or a storm — and a substantial fraction of the occupation (theorists, HEP analysts, modelers) never touches hardware at all.
Certification preferred, not legally required A 5 is the weighted reality: the PhD is a hiring credential with no licensing statute behind it, and only the medical physics slice carries ABR board certification plus state radiation-machine registration where a treatment-plan error lands on a named individual — that minority is what lifts the score off the floor rather than any protection covering the beamline or condensed-matter physicist.
Exists to be accountable for ambiguous calls A 14 is earned by calls that no procedure covers: declaring a measurement publishable, setting the systematic error budget that determines whether a discovery claim survives, deciding a novel apparatus is safe to energise, and in the clinical case approving a dose distribution that will be delivered to a patient — decisions where being wrong costs a retraction, a destroyed multi-million-dollar instrument, or a radiation injury.
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.
Your task mix speaks to task resistance (11/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (5/20) is whether the law requires a licensed human to sign. Trust premium (7/20) is whether buyers specifically pay for a person. Judgment and accountability (14/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 26 of this occupation's 46 points (57%).
Embodiment (9/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
No occupation passed every test: close enough to physicists on skills and subject matter, at least 10 points more resistant, no big jump in training, no new licence, no pay cut, and not shrinking on its own. That happens for 223 of the 654 occupations here that aren't SAFE, and it is worth stating plainly rather than leaving the section off.
The usual reason is that exposure travels with the skill profile. The jobs most similar to yours tend to be exposed for the same reasons yours is, so the near neighbours don't clear the gap — and the ones that do are a different kind of work, not a transfer of what you already know. Read that as a limit of this method, not a verdict that you're stuck: it only compares whole occupations, and it cannot see specialisation, industry, or anything you'd bring that isn't in a federal skill survey.
Here is that claim on your own job rather than in the abstract. These are the three occupations closest to this one by skill and subject matter — the places the work would most naturally transfer — with what the register scores them:
That is the whole problem in three lines. The nearest work is not meaningfully safer, so there is no move here that trades a similar skill set for a better verdict. This is not us running out of ideas — it is what the neighbourhood looks like.
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 63/100, still EXPOSED.
State licensure of medical physicists (bills already enacted in TX, FL, NY, HI and pushed by AAPM) extended to more states, plus an ACR–AAPM practice parameter or NRC rule requiring a named Qualified Medical Physicist to personally review and sign AI-generated treatment plans, auto-contours and machine QA reports. Also watch NRC/state radiation-safety rules naming an individual Radiation Safety Officer accountable for AI-assisted dosimetry.
Narrow route only: classified and export-controlled work (DOE Q clearance, ITAR-covered defense physics) where the buyer requires a cleared human to hold and interpret the data, and cloud AI tools are prohibited on the network. This is a security requirement rather than a preference for human authorship, so it is a floor rather than a premium.
Genuine two-tier structure: if literature synthesis, fitting, code, and draft text are absorbed, the residual week is experiment design never run before, discriminating true anomaly from instrument artifact, and choosing which systematic to chase — the tier where verification cost exceeds generation cost. Measurable signal: national-lab staffing shifting from analysis FTEs toward instrumentation/commissioning roles at facilities like LCLS-II, EIC, or fusion pilot programs.
Funder and journal rules that place non-delegable responsibility on a named human for AI-produced content: NSF's and NIH's prohibition on generative AI in peer review, NIH's 2025 policy limiting AI-generated applications, and DOE lab research-integrity procedures requiring a named principal investigator to attest that simulation/analysis outputs were independently verified. If audit-with-personal-attestation becomes standard for computational results, the calls under ambiguity concentrate in the PI role.
Growth in the share of physicist headcount attached to apparatus that cannot be simulated away — beamline commissioning, dilution-refrigerator and qubit-fab work in the national quantum initiative centers, ITER/SPARC diagnostics, detector installation. This is a composition shift within the SOC, not a change in the work itself; it only raises the score if hiring actually tilts that way.
The limit. Outside medical physics and radiation safety, there is no licensure body that could plausibly require a physicist's personal signature — no client pays for a human-derived cross-section or human-written simulation as such. Liability shield and trust premium for the theory/computational majority of this SOC have no realistic route upward; the durable gains sit in task-mix and in apparatus-attached roles.
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 1,520 | $171,580 +0% |
| Chicago-Naperville-Elgin, IL-IN | 1,510 | $137,140 -20% |
| New York-Newark-Jersey City, NY-NJ | 1,070 | $181,970 +6% |
| Boston-Cambridge-Newton, MA-NH | 1,060 | $106,000 -38% |
| Boulder, CO | 860 | $104,580 -39% |
| Los Angeles-Long Beach-Anaheim, CA | 550 | $160,270 -7% |
| San Jose-Sunnyvale-Santa Clara, CA | 440 | $173,450 +1% |
| Albuquerque, NM | 330 | $165,340 -4% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 100 | $272,480 +58% |
| Tampa-St. Petersburg-Clearwater, FL | 70 | $266,260 +55% |
| Cleveland, OH | 240 | $239,360 +39% |
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 46. 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.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.