AI-proof jobs: which work is least exposed to AI, and why "proof" is the wrong word

Jobs less exposed to AI, with their numbers

Two figures for each job. AI exposure is Anthropic's "observed exposure" (March 2026): how much of the job's work people actually do with its AI model, Claude, in a work setting, with full weight when AI does a task and half when it helps. Projected change is the Bureau of Labor Statistics projection for 2025 to 2035, when all jobs together grow 3.5%.

AI exposure for every row is from Anthropic's published file of scores by occupation. Anthropic's own examples of jobs with no measurable exposure include cooks, motorcycle mechanics, lifeguards, bartenders and dishwashers.

What these jobs share: the work happens with your hands, on site, or with a patient in front of you, and someone has to be there and be responsible for it.

Why "AI-proof" is the wrong word

• A 0% is not a guarantee. Anthropic says these jobs score zero because their tasks "appeared too infrequently in our data to meet the minimum threshold." It measures use of one company's chatbot, not robots, and not what AI might do in ten years.
• Exposure and job growth do not move one for one. See the table below: some heavily exposed jobs are projected to grow fast. Anthropic found only a mild link overall: for every 10 points more exposure, the BLS growth projection was 0.6 points lower.
• Projections are estimates. BLS warns that the precision of its figures "does not account for the inherent uncertainty of predicting long-term changes in the labor market."
• Usage data measures how people use AI tools, not job losses. Anthropic found "no systematic increase in unemployment for highly exposed workers since late 2022."
• Nobody knows the answer for certain. A list that calls jobs "AI-proof" is claiming more than anyone can know. "Less exposed today" is the honest version.

High exposure does not always mean decline

Programmers and customer service representatives are both highly exposed and projected to shrink, and BLS names automation as a reason for each. Data scientists and security analysts are also highly exposed, and among the fastest growing jobs. The difference is whether demand for the work is growing faster than AI speeds it up. Read more in will AI replace software engineers.

What jobs are safe from AI in your own field?

You may not need to change careers to be less exposed. Inside almost every job, some tasks show heavy AI use and others almost none. Our occupation pages split them out:

• Will AI replace doctors? Family doctors score 0%, radiologists 20%.
• Will AI replace teachers? Kindergarten teachers score 0%, high school teachers 29%.
• Will AI replace lawyers? Lawyers score 17%, paralegals 29%.
• Will AI replace accountants? Accountants score 35%.

For how to judge your own job by its tasks, start with will AI take my job.

What to do with this

If you are choosing a field, use both columns in the first table, and open the BLS page for each job to see the training and licence it needs.

If you are staying where you are, move the less exposed parts of your work to the top of your resume: the hands-on work, the decisions you were responsible for, the people you served or led. Then run the resume check on this page, which is free and needs no account. It reads your actual resume and shows how your experience comes across to an employer's software and to a recruiter. It does not score your job's AI risk; it shows whether your strongest work reads that way.