Will AI take your job? A new 37-country poll shows what people fear—and what it can’t tell you

In 34 of 37 countries, more people expect AI to reduce jobs than create them. That is a view about the future of national job markets—not a prediction about your employer or your next shift.

You ask AI to draft the meeting notes. It does the job in seconds. The relief lasts until the next thought: if it can do this, could it do the rest of my work too?

A fresh Pew Research Center survey puts that fear on a global map. The center surveyed more than 42,000 adults in 36 countries between February and May 2026, then added separate U.S. samples for a 37-country comparison. In 34 of those countries, more people expected AI to lead to fewer jobs than more. In Australia, South Korea and the United States, about seven in ten adults or more expected AI to reduce jobs over the next 20 years.

That is a striking number, but it measures what people think will happen—not jobs already lost, a company's plan, or your personal odds. The same report found substantial uncertainty in some countries. A public worry can be real without being a job-loss forecast.

There is a different way to measure exposure. The International Labour Organization examined tasks within occupations and estimated that one in four workers worldwide is in a job with some degree of generative-AI exposure. The ILO says most jobs are more likely to be transformed than made redundant. “Exposure” means that AI could potentially perform part of a task; it does not mean the task has been automated or that a worker will be laid off. In 2026, the ILO cautioned that exposure indicators alone cannot predict employment outcomes.

So a job title is a poor shortcut for judging your own situation. Break the role into actual recurring tasks instead: where AI might help draft or organize, where a person must verify the output, and where judgment, relationships, physical presence or accountability matter. Try the prompt below with generic task descriptions only.

Help me map tasks in my role; do not predict whether I will keep my job. I will describe generic tasks only—no employer secrets, customer data, names, internal documents, or personal information. For each task, separate: (1) where AI may help draft or organize, (2) where it could automate part of a repeatable step but needs human checking, and (3) where human judgment, relationships, physical presence, or accountability matter. Do not infer risk from my job title alone. Then suggest one low-risk 15-minute test and how to check its output for errors.

My recurring tasks, described generically: [list 3–5 tasks]

At work, check whether your employer approves the tool before using it. Do not paste confidential, customer or personal information into an unapproved service. Test one low-risk task and compare the time, errors and correction work with your usual method. One good result tells you about one task—not whether your whole job is safe or doomed.

The useful question is smaller: what part of your actual work can AI assist, and where must a person remain responsible?

Pew measured adults’ expectations, not actual layoffs. ILO exposure estimates describe task potential, not individual job-loss probabilities; outcomes vary by occupation, country and employer.

Sources: Pew Research Center, global views of AI and jobs (2026); International Labour Organization, Generative AI and Jobs: A 2025 Update; ILO, what AI exposure indicators can and cannot show (2026).