Back to quiz

How we score jobs

Our quiz gives each job a score from 0 to 100. The higher the score, the more AI is likely to change that job. Here's how we figure it out.

Where our data comes from

We use two main sources:

  • Karpathy's AI Exposure Index — Andrej Karpathy (co-founder of OpenAI) built a tool that scores 342 jobs from the Bureau of Labor Statistics on a 0-10 scale. An AI model looks at what each job involves and rates how much of it could be done or changed by AI. We scaled this to 0-100 for our quiz.
  • Anthropic's Labor Market Research — Anthropic (makers of Claude) studied real-world AI usage to see which jobs actually use AI the most right now. They found that some jobs that could theoretically be automated (like programming) actually see more demand, not less.

What the scores mean

Score Rating What it means
80-100 Very Bad AI can already do a big part of this job. Major changes are coming fast.
60-79 Bad AI is changing a lot about this job. You'll need to learn new tools to stay competitive.
40-59 Okay Some parts of the job are changing, but a lot of it still needs a real person.
20-39 Good AI doesn't affect much of this job. You're in a solid position.
0-19 Great AI has almost no impact. This job needs hands-on, in-person human skills.

Important things to know

  • A high score doesn't mean you'll lose your job. Software developers score 85/100 but there are more developer jobs than ever. A high score means the job is changing a lot, not necessarily disappearing.
  • These are estimates, not predictions. Nobody knows exactly what will happen. We're using the best data available, but the future is uncertain.
  • Your specific role matters. Two people with the same job title can have very different day-to-day work. Your actual risk depends on what you specifically do.
  • AI as a tool is different from AI as a replacement. Many jobs are being helped by AI (augmented) rather than replaced. Doctors using AI to read scans faster is very different from AI replacing doctors.

The key idea

Karpathy uses a simple rule of thumb: if your job can be done entirely from home on a computer, it scores 7+ out of 10. Jobs that need you to be physically present, work with your hands, or deal with unpredictable real-world situations score much lower.

Anthropic's research adds a reality check: just because AI could do something doesn't mean it actually is. Their study found that only about a third of theoretically automatable computer work is actually being done by AI right now.

What's the "Oxford risk" number?

This comes from a famous 2013 study by Carl Frey and Michael Osborne at Oxford University. They gave 702 US jobs a probability between 0% and 100% for how likely that job is to be computerised. A score of 98% means their model thinks the job is very likely to be automated. A score of 1% means it's very unlikely.

Important: this study was done before modern AI like ChatGPT existed. Some of its predictions have held up, others haven't. We show it alongside our own score so you can compare. Our score accounts for more recent data.

Where do the salary numbers come from?

  • US salaries come from the Bureau of Labor Statistics (BLS) Occupational Employment and Wage Statistics. These are median annual wages.
  • UK salaries come from the Office for National Statistics (ONS) Annual Survey of Hours and Earnings (ASHE). These are median annual earnings.

All our data sources

🤖 Take the Quiz