How the AI Survival Score works

This is an estimate, not a prediction and not career advice. Nobody knows exactly how AI will change any one person's job. The score is a structured way to think about your own exposure and what you can do about it.

Tasks, not job titles

AI doesn't replace whole jobs; it takes on tasks. Two people with the same title can spend their weeks very differently, so the model starts from what you actually do. It maps your answers onto 16 task clusters, from data entry and drafting to care, negotiation and hands-on field work, and 15 role families.

The five things we measure

• Task safety: how hard the tasks that fill your week are for current AI to do or speed up.
• AI adoption: how often and how widely you already use AI tools.
• Learning: how recently you learned something new, how much time you invest in learning, and how you respond when tools change.
• Resilience: your network, income outside your main job, and the human parts of your work that AI finds hard to take over.
• Market: our estimate of the demand trend for your role family as AI spreads.

The model combines these into a score between 3 and 97. We never show 0 or 100, because nobody is certainly safe or certainly replaced. The exact weighting is proprietary and is reviewed as AI capabilities change.

Consistent and explainable

Your score comes from our scoring model, not from a chatbot's opinion. The same answers always give the same score, and your result page shows how you did on each of the five factors, so you can see what is driving your number and what would move it.

The four archetypes

The model also places you on two axes: how exposed your tasks are, and how fast you are adapting. Together they give one of four archetypes:

• The Irreplaceable: less exposed work, adapting fast.
• The Centaur: exposed work, but fused with AI.
• The Comfortable: less exposed work, not adapting yet.
• The Endangered: exposed work, not adapting yet.

What to learn next

Your result includes a short list of skills to learn, chosen for your archetype and your most exposed tasks, plus three practical moves for your role. You can explore the same thinking task by task on ourtask pages.

The research behind it

We built the model by working through published research on how exposed different kinds of work are to generative AI, and on how people actually use AI at work, then making our own calibration choices. In particular:

• Eloundou, Manning, Mishkin and Rock, "GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models" (OpenAI and the University of Pennsylvania, 2023), which rated task-level exposure across occupations.
• The International Labour Organization's research on generative AI and jobs (Gmyrek, Berg and Bescond, 2023, and later updates), which looked at exposure by occupation and the difference between automation and augmentation.
• The Anthropic Economic Index, which reports which tasks people actually use an AI assistant for.

These sources informed our work; they did not produce our numbers and do not endorse them. For a plain-English summary, read what the research really says about AI and jobs.

Limitations

• Twelve multiple-choice questions cannot capture a whole job. Two people with the same title can do very different work.
• Any model of the future of work is an estimate. AI capabilities change quickly, and we update ours as evidence improves.
• Exposure is not the same as job loss. A task AI can do may still be done by people because of cost, trust, regulation or preference.
• The score does not know your country, your employer, your finances or your health, and it does not try to.
• The written text on the result page was drafted with AI help and reviewed for tone. It never changes your score.

If you think we've got something wrong for your kind of work, we'd genuinely like to hear why. See thecontact page.