How we score careers

What we score

All 831 occupations in the US Bureau of Labor Statistics projections, across 22 kinds of work, from healthcare and the skilled trades to law, finance, technology and the arts. That is every occupation in the BLS projections, not a hand-picked list.

Where the evidence comes from

Published datasets, used as they were released: the BLS 2025–35 projections and its new AI exposure categories (August 2026), O*NET 31.0 (August 2026), Microsoft's study of 200,000 Copilot conversations (2025), Anthropic's observed exposure (March 2026) and robot exposure (September 2026) data, the OpenAI and University of Pennsylvania exposure ratings (Science, 2024), and Felten, Raj and Seamans (2023). The headline findings around them come from the ILO, IMF, OECD, World Economic Forum, Stanford, PwC, McKinsey, Goldman Sachs, the New York Fed, Yale and Indeed. We read the original reports, not summaries of them.

What goes into the score

• How much of the work AI can do (40%). Four independent measures, each turned into a ranking and averaged: two that test what AI could do in theory, and two that measure what people actually use AI for. This is similar to the method the Bureau of Labor Statistics used for its own AI exposure categories, which combined these four and one more.
• The four human superpowers (30%): emotional intelligence, creative vision, physical dexterity and ethical judgement, measured from O*NET ratings of the skills, abilities and work activities a job needs. We count each career's two strongest. Physical dexterity counts for less where robots can already do the physical tasks.
• Whether employers are hiring (30%): projected growth from 2025 to 2035 and openings each year, from the Bureau of Labor Statistics.
• The first rung. We flag careers that are highly exposed and usually entered straight from education, because research in 2026 finds young people are being hired less into jobs like these. The flag sits beside the score rather than inside it.

One worked example: physical therapists

• Daily work: AI exposure is in the middle (BLS rates it High). People already use AI to explain medical information and handle paperwork, but examining and treating patients sees little AI use.
• Human strengths: very high on emotional intelligence (86) and ethical judgement (85). Robots can do almost none of the physical work today.
• Hiring: BLS projects 11.9% growth by 2035, much faster than the 3.5% average, with about 13,400 openings a year.
• Result: 76, in the "Grows with AI" group. AI will change parts of the job; demand is still strong.

Figures from the 2026 edition of the Career Explorer.

What changed in the 2026 edition

• We now score all 831 US occupations from published data. The 2025 edition scored 35 careers.
• The score is now a ranking from 1 to 99, so it compares a career with every other career. Because of that, most numbers look lower than in the 2025 edition. Compare careers with each other, not with last year's numbers.
• Some familiar careers moved down the ranking. Teachers, counsellors, doctors and many office, writing and computing jobs rank lower, because new evidence on how people actually use AI shows more of their tasks can be helped by it, and some are projected to grow slowly. Hands-on careers in healthcare, the trades and public safety rank near the top.
• We added BLS's own AI exposure category, robot exposure, and a map that separates careers that are exposed but still growing from those that are changing shape.

What a score cannot tell you

• It describes a career on average, not a particular young person. Interest, ability and effort matter more than any score.
• Exposure to AI is not a forecast of job loss. Many highly exposed careers are still growing, because AI changes how the work is done.
• It reflects published research up to the date shown. AI is changing quickly, so we date every edition and explain what changed.
• It is US data. A career can look different in another country, which is why paid Deep-Dives are country by country.
• It cannot tell you whether someone will enjoy the work. Only trying it can.

Our rules

• Every figure comes from a published dataset or report, and we link to the source.
• No machine-made statistics. Mr Thoob's plain-words answers are told to use only the figures already on the page, and every career page shows the source data.
• Every edition is dated, and changes are explained. The full table is free to download.
• No sponsors and no paid placements.