SAFE
Psychiatry is largely conversation and prescribing, so the documentation layer — intake summaries, progress notes, medication-interaction checks, prior-authorization letters, screening-scale scoring — is already being absorbed by AI. What does not move is the controlled-substance prescription, the involuntary-hold decision, the suicide-risk call made on incomplete information, and the patient who will only disclose to a physician they trust. Telepsychiatry lowers the embodiment moat, but licensure, DEA authority, and personal malpractice exposure remain the hard floor.
Headcount grew steadily across the period.
Median pay $239,200 → $281,870 -5.7% 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
+6.1% 27,100 → 28,800 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +6.1% 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.
~900 openings a year on average, including replacing people who leave.
DoctorPhysicianPsychiatristPsychoanalystNeuropsychiatristAdult PsychiatristBehavioral AnalystChild PsychiatristStaff PsychiatristMedical Doctor (MD)Prison PsychiatristGeneral PsychiatristPsychiatry PhysicianBehavioral SpecialistClinical PsychiatristPsychiatric PhysicianAddiction PsychiatristInpatient PsychiatristPediatric PsychiatristConsulting PsychiatristMental Health PhysicianOutpatient PsychiatristGeneral Adult PsychiatristAdult Inpatient Psychiatrist
Holding it up: liability shield . Weakest point: embodiment .
Tasks largely resist digitisation A 45-minute diagnostic evaluation where the differential turns on flat affect, pressured speech, and whether the patient is minimizing — plus titrating clozapine against ANC counts and negotiating a treatment plan with a family that disagrees with it — sits at 14 rather than 18 because the surrounding PHQ-9/GAD-7 scoring, note generation, and interaction screening are already machine work.
Some physical or field component An 8 reflects the physical minimum that remains: catatonia and EPS exams, AIMS scoring for tardive dyskinesia, IM long-acting injectable administration, ECT supervision under anesthesia, and being bodily present on a locked inpatient unit during a code — but a large share of outpatient med management now happens over video from a home office.
Licensed human required and personally liable A 19 is set by the DEA registration required to write for buprenorphine, stimulants, and benzodiazepines, state medical licensure, and the fact that a psychiatrist's signature on a 5150/72-hour hold and on every prescription is the personal liability anchor in a wrongful-death or Tarasoff duty-to-warn suit.
Exists to be accountable for ambiguous calls Deciding to commit someone against their will, discontinuing lithium in a pregnant patient, or judging whether a threat to a third party breaches confidentiality are calls made on partial and often deliberately distorted information, with no protocol that resolves them and the psychiatrist's name on the order.
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 (14/20 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 (19/20) is whether the law requires a licensed human to sign. Trust premium (18/20) is whether buyers specifically pay for a person. Judgment and accountability (17/20) 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 54 of this occupation's 76 points (71%).
Embodiment (8/20) 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.
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 84/100, still SAFE.
Task-mix shift: as ambient scribes (Abridge, Nuance DAX), automated PHQ-9/GAD-7 scoring and prior-auth letter generation absorb the documentation tier, the residual job concentrates on diagnostic ambiguity, treatment-resistant cases, suicide-risk calls on incomplete history, and clozapine/lithium/MAOI management — tiers with no clean ground truth. Also if boards or payers bar autonomous AI from psychiatric diagnosis, as some state AI-in-mental-health bills (Illinois HB 1806, 2025, restricting AI therapy) do
DEA's post-pandemic telemedicine controlled-substance rulemaking finalizing a requirement that Schedule II prescriptions (stimulants, ketamine) carry an in-person or synchronous physician evaluation, plus state PDMP rules naming the prescriber personally accountable for AI-suggested regimens; also state statutes (e.g. involuntary-commitment codes) that name a physician or psychiatrist as the only party who may certify an emergency hold
Malpractice suits or FTC actions following harm from consumer AI chatbots (already filed against Character.AI) hardening a market norm that psychiatric care be physician-delivered; cash-pay concierge psychiatry where the fee is explicitly for a named physician's continuity
Expansion of statutory duty-to-warn/Tarasoff-type liability and state rules making the psychiatrist the accountable party for risk-stratification outputs generated by AI triage tools, so the consequential call remains personally attributable
The limit. Already at 76 with liability_shield near maximum; realistic headroom is a few points, mostly from task-mix concentration. Embodiment has no plausible upward route — telepsychiatry pushes it down, not up.
| New York-Newark-Jersey City, NY-NJ | 4,180 | $249,130 -12% |
| Los Angeles-Long Beach-Anaheim, CA | 1,610 | $370,510 +31% |
| Boston-Cambridge-Newton, MA-NH | 760 | — |
| Miami-Fort Lauderdale-West Palm Beach, FL | 720 | $403,430 +43% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 640 | $279,410 -1% |
| Chicago-Naperville-Elgin, IL-IN | 600 | $258,750 -8% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 550 | $246,990 -12% |
| Riverside-San Bernardino-Ontario, CA | 490 | $364,530 +29% |
| Springfield, IL | 40 | $413,890 +47% |
| Burlington-South Burlington, VT | 50 | $409,950 +45% |
| Miami-Fort Lauderdale-West Palm Beach, FL | 720 | $403,430 +43% |
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 76. 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.