The 2026 AI Job Displacement Index
100 Careers Ranked by What Protects Your Income

Most AI job-risk rankings answer the wrong question. They measure exposure — how many of a job's tasks an AI can perform — and on that score, programmers, translators, and writers top almost every list. But workers need a different answer: not what AI can touch, but whether it actually costs them the job. Those are not the same thing. In the sectors most exposed to AI, employment has slipped only about 1% since late 2022 — even as average wages in those same sectors rose faster than the national average — according to the Federal Reserve Bank of Dallas. The job losses that have appeared are concentrated among entry-level workers, while experienced ones are seeing AI complement their work. Globally, the World Economic Forum projects AI and related trends will displace about 92 million roles by 2030 while creating 170 million — a net gain, but an uneven reshuffling of which jobs exist, where, and for whom.

The 2026 AI Job Displacement Index ranks 100 careers by that second measure: resistance to actual job loss, scored across four dimensions the exposure rankings leave out.

Why exposure rankings are useful — but incomplete

The dominant method for predicting automation risk is task decomposition: break a job into its constituent tasks and ask which ones a machine could perform. The approach traces to a 2013 University of Oxford study by Frey and Osborne — written before large language models existed — which assigned telemarketers a 99% automation probability and surgeons under 1%. It got telemarketers right. It badly missed how deeply AI would later embed itself in cognitive work like diagnostics, drafting, and analysis. The framework tends to overstate risk for physical jobs and understate it for knowledge jobs.

The newest exposure studies are far more sophisticated — and worth taking seriously — but they still measure task overlap, not job loss. The Anthropic Economic Index, built on millions of real Claude conversations rather than hypothetical task lists, finds AI use concentrated in software development and technical writing, and higher for mid-to-high-wage roles like computer programmers and data scientists; roughly a third of occupations now see AI used in at least a quarter of their tasks. Microsoft Research, analyzing about 200,000 Copilot conversations across 785 occupations, ranked interpreters and translators highest for "AI applicability" — about 98% of their work activities overlap with tasks Copilot is asked to do — with writers, historians, and customer service close behind. Crucially, Microsoft's own researchers stressed that a high applicability score does not mean the job will be replaced.

That distinction is the whole game, and the most rigorous index in the field now builds on it. Tufts University's American AI Jobs Risk Index deliberately measures vulnerability to job loss, not merely exposure across 784 occupations. It projects roughly 9.3 million U.S. jobs at risk of displacement within two to five years, concentrated in 33 "tipping point" occupations — about 4.9 million workers — whose risk swings from under 10% to over 40% depending on how fast adoption accelerates.

Every exposure ranking tells you what AI can touch. None tells you whether the worker actually gets displaced — because displacement is gated by forces that live outside the task list: whether a licensed human must be legally accountable, whether the work requires a body in the room, and whether the field is already short of people. That gap is what this Index scores.

The four dimensions that actually predict displacement

FutureJobRisk's model isolates four dimensions that consistently separate the jobs AI displaces from the jobs it merely assists. Each career on the Index is scored against all four.

Physical Presence Required

Can the work be done remotely by a system without a body? Jobs that demand a licensed, insured, physically present human — the plumber under the sink, the nurse at the bedside — sit at the top of the Index. This is the single strongest protective factor in the model.

Unpredictable Human Interaction

AI can simulate empathy. What it struggles with is contexts where the other person knows they're dealing with a machine, and where the work depends on building trust over time, navigating conflict, and reading emotional states as they shift.

Adaptive Judgment in Novel Environments

A plumber never meets the same pipe twice; a therapist never sees the same session twice. Real-time judgment in unrepeatable, unstructured situations resists automation in a way that pattern recognition on clean, structured data does not.

Regulatory and Licensing Moats

A state license requires a human to be legally responsible. AI can assist a doctor, a lawyer, or an electrician — but under current frameworks it cannot hold the license, carry the liability, or answer to a regulatory board. These moats tend to widen under pressure, not shrink.

The 2026 AI Job Displacement Index

One hundred careers, scored for their 5–10-year resistance to AI displacement as of mid-2026. A higher score means stronger structural protection against full displacement — not that the job will be unchanged. Search, filter by sector or verdict, or tap any column to sort.

The biggest exposure-vs-displacement gaps

This is where the Index earns its keep. The jobs that look safest on an exposure chart are sometimes the most displaced — and several of the "most exposed" jobs are among the best protected.

Translators sit near the top of every exposure study — and near the bottom of ours (39, High Risk). Here exposure and displacement point the same way: the task overlap is real and the protective dimensions are weak. This is what genuine high risk looks like, and the exposure rankings get it right.

Nurses are the inverse. Their AI task overlap is in the single digits, and FutureJobRisk scores the registered nurse role at 94. But the protection isn't only that AI can't do the tasks — it's that hospitals are buying AI precisely because they cannot hire enough nurses. Demand, physical presence, and licensing all reinforce each other. (We unpack this in Will AI Replace Nurses?.)

Doctors are safe — but not for the reason people assume. Complexity isn't the moat; presence and liability are. AI diagnostic tools already match or beat specialists at detecting certain conditions. A primary care doctor scores 82 because a licensed human must sign the order, carry the malpractice exposure, and sit across from the patient — not because the cognitive work is beyond AI.

Which is why being a "doctor" isn't a uniform shield. Radiologists (65) and pathologists (61) are physicians whose core work is pattern recognition on structured images and slides — exactly what AI does well — so they score far below surgeons (85) and primary care. Inside one profession, physical contact is the dividing line.

The sharpest gap of all runs through tech — and the Dallas Fed data backs it. A senior software engineer scores 68; a junior one scores 45. The Fed found AI is hitting entry-level workers in exposed sectors hardest while raising pay for experienced ones — exactly the split the Index captures. The same junior-senior divide shows up in graphic design (senior 64, junior 38) and UX. Rankings that lump "software engineer" into one bucket miss the most important fact about the field.

What the Index says by sector

Healthcare is the most resistant major sector — not because medicine is complex, but because liability, licensing, and physical presence are mandated at once. The split is internal: hands-on roles (physical and occupational therapists, surgeons, nurses) are extremely safe; structured-data roles (radiology, pathology, medical coding) are under real pressure.

Skilled trades are the most underrated category on the list. Every plumbing job and electrical panel is different, and the capital cost of a robot that handles that variability still runs well above a licensed tradesperson over a five-year horizon. A long-running skilled-trades labor shortage only deepens the moat: the BLS projects electrician employment to grow 9% from 2024 to 2034 — much faster than the average for all occupations — with about 81,000 openings a year, many of them driven by workers retiring or otherwise leaving the trade.

Technology is the most internally divided sector: cybersecurity analysts (77) and AI/ML engineers (76) sit near the top while junior developers and IT support face genuine pressure.

Law and finance are more exposed than their practitioners tend to believe, because so much of the day-to-day work is document-intensive. The survivors are built around judgment, advocacy, and relationships — which is why a litigator (78) far outscores a paralegal (55), and why financial advisors (76) who shifted from managing portfolios to coaching behavior held their ground while accountants (68) feel the document-automation squeeze.

What to do with your score

• 80+ — Strong structural protection. The risk isn't displacement; it's commoditization of the lower rungs of your field. Stay fluent in the AI tools in your profession and use them to raise your output.
• 60–79 — You likely have a window before serious structural pressure. This is the time to move up the value stack or make a lateral move toward a higher-scoring adjacent role.
• Below 60 — The disruption isn't coming; it's underway. Roles built on pattern recognition, document processing, or routine execution are already seeing reduced hiring.