SAFE
The core work — sitting with a couple mid-conflict, tracking what each partner isn't saying, interrupting a destructive pattern in real time, and holding a family's trust across months — is exactly what language models cannot do, because the product is a relationship with an accountable human, not a transcript. LMFT licensure, mandatory reporting duties, and insurance credentialing require a named clinician to own the treatment plan and the risk assessment. Chatbots will absorb the low-acuity, low-cost end of counseling demand and the paperwork layer (notes, treatment-plan drafts, intake summaries), which compresses fees more than headcount.
Dipped in 2020, then grew past where it started.
Median pay $49,610 → $66,940 +7.9% 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
+12.6%
Percentage only. The projection counts a different population from the 66,740 above — it includes self-employed workers, which for this occupation is most of them, so the two headcounts are not comparable.
Hard to automate, and growing
The work resists current AI and the BLS projects +12.6% 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.
~7,700 openings a year on average, including replacing people who leave.
ClinicianCounselorTherapistCase ManagerDrug CounselorPlay TherapistElder CounselorGrief CounselorGroup CounselorBehavior AnalystFamily CounselorFamily TherapistCouples TherapistBehavior TherapistBehavioral AnalystClinical CounselorClinical TherapistLicensed ClinicianLicensed CounselorLicensed TherapistMarriage CounselorMarriage TherapistQuitline CounselorAddiction Counselor
Holding it up: trust premium . Weakest point: embodiment .
Tasks largely resist digitisation Genogram work, structural interventions like enactments where you deliberately let a couple fight in front of you and redirect it mid-sentence, and de-escalating a suicidal adolescent while the parents are in the room are sequential judgment calls with no scriptable next step — the 15 rather than 18 reflects that intake screening, PHQ-9/GAD-7 administration, treatment-plan boilerplate, progress-note drafting, and psychoeducation handouts are already being generated for you.
Some physical or field component The 9 sits above desk-only because you read shifted posture, a partner's withdrawal to the far end of the couch, alcohol on someone's breath, and bruises on a child, and you manage physical room dynamics — seating, who sits where, separating a couple — plus school meetings, court appearances, and occasional home visits; it stops at 9 because telehealth-delivered MFT is reimbursed and clinically established, so the body in the room is an advantage, not a requirement.
Licensed human required and personally liable State LMFT licensure with 3,000-odd supervised hours, a named clinician on every insurance claim and treatment plan, mandatory reporting duties under state child- and elder-abuse statutes, Tarasoff-line duty-to-warn exposure, and board complaints that can end the career mean the risk attaches to a person; it is 15 not 19 because scope of practice overlaps heavily with LPCs, LCSWs, and psychologists, so the license protects the title more than it monopolizes the work.
Exists to be accountable for ambiguous calls You decide whether reported conflict crosses into intimate-partner violence that makes conjoint therapy unsafe, whether a disclosure triggers a report that will blow up the family you are treating, when to break confidentiality for imminent risk, and when to terminate or refer — calls made in the room without a second opinion; 16 rather than 19 because you work inside diagnostic criteria, agency protocols, and utilization review rather than setting the terms yourself.
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 (15/20 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 (15/20) is whether the law requires a licensed human to sign. Trust premium (19/20) is whether buyers specifically pay for a person. Judgment and accountability (16/20) 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 50 of this occupation's 74 points (68%).
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.
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 83/100, still SAFE.
Extension of the Illinois Wellness and Oversight for Psychological Resources Act (HB1806, 2025) model — which bars AI from delivering therapy or making therapeutic decisions absent a licensed clinician — to more states, plus Nevada AB406 and Utah HB452-style rules; if boards additionally require a named LMFT to review and countersign any AI-generated treatment plan, risk assessment, or progress note used for insurance billing, the shield tightens
Task-mix shift: chatbots absorb low-acuity single-person coaching and the documentation layer, leaving the remaining caseload concentrated in high-acuity work AI is worst at — multi-party conflict in the room, domestic-violence screening where a joint session is contraindicated, active infidelity disclosure, family systems with a child protection concern. This is a genuine two-tier occupation and the judgment tier is what remains
Malpractice carriers (CPH & Associates, HPSO) writing exclusions that void coverage unless a licensed clinician personally conducts suicide/violence risk assessment and safety planning, making the licensed signature non-delegable in practice as well as in statute
Courts and child-welfare agencies continuing to require a licensed human for custody evaluations, reunification recommendations, and court-ordered family therapy reports — plus state rules explicitly barring AI-generated input into mandated-reporter determinations, so the consequential calls (report or not, break confidentiality or not, hospitalize or not) stay owned by a named clinician
The limit. trust_premium at 19 is effectively maxed — buyers already pay specifically for a human relationship and there is no realistic route higher. Embodiment cannot rise; presence in a room is not the same as physical manipulation of an unpredictable environment. The real exposure here is fee compression and payer substitution of chatbots for low-acuity covered benefits, which the score does not capture and no lever above prevents.
| Los Angeles-Long Beach-Anaheim, CA | 13,200 | $71,110 +6% |
| San Diego-Chula Vista-Carlsbad, CA | 5,350 | $49,610 -26% |
| New York-Newark-Jersey City, NY-NJ | 2,800 | $91,130 +36% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 2,110 | $78,640 +17% |
| Sacramento-Roseville-Folsom, CA | 1,380 | $75,030 +12% |
| San Jose-Sunnyvale-Santa Clara, CA | 1,180 | $95,110 +42% |
| Salt Lake City-Murray, UT | 890 | $86,550 +29% |
| Provo-Orem-Lehi, UT | 730 | $88,810 +33% |
| Urban Honolulu, HI | 190 | $141,960 +112% |
| Greensboro-High Point, NC | 90 | $109,200 +63% |
| Eugene-Springfield, OR | 130 | $102,910 +54% |
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 74. 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.