The headline number
In 34 of the 37 countries Pew Research Center surveyed this year, more people expect artificial intelligence to reduce the number of jobs than to increase it. The centre published the results on Thursday, drawing on interviews with 42,151 people conducted between 8 February and 13 May 2026.
A global median of 46% said AI will lead to fewer jobs over the next twenty years. That figure conceals more than it reveals, because the variation between countries is where the survey gets interesting.
Wealth predicts worry
Across 18 high-income countries, a median of 55% expect AI to reduce employment. Across 18 middle-income countries, the figure is 36%.
That is the opposite of what a simple exposure argument would predict. If the fear tracked vulnerability to automation, the countries with more routine manufacturing and services work would be the anxious ones. They are not. The anxious ones are the countries where AI tools are most widely available, most discussed in national media, and most present in white-collar work.
The United States sits near the top: 71% of American respondents expect AI to lead to fewer jobs, up seven points in two years. Only Australia and South Korea are higher, at 76% each.
Concern is not the same as opposition
The sentiment data is more mixed than the employment question suggests. A median of 37% say they feel more concerned than excited about AI, while 41% report feeling concern and excitement in equal measure. Israel is the only country surveyed where more people say they are primarily excited than primarily concerned.
On inequality, respondents lean towards expecting AI to widen wealth gaps, but Pew is careful about the strength of that finding: relatively large shares in many countries are either unsure or say AI will not affect inequality much. This is a question the public has not settled with itself.
The age pattern has shifted
One change from earlier rounds is generational. Younger adults, aged 18 to 34, have historically been more sanguine about AI than older respondents. In this round they increasingly share the same worries, particularly about job displacement and inequality. The cohort most likely to use these tools daily is not the cohort most confident about what they will do to work.
What it does and does not tell you
A survey measures expectations, not outcomes. Nothing here establishes what AI will do to employment; it establishes what large publics across 37 countries currently believe, and that belief is itself politically consequential. Legislators facing votes on AI rules are looking at electorates in which pessimism about jobs is the majority view almost everywhere, and is strongest in the wealthy democracies that write most of the rules.
The fieldwork closed in May. Anything that has happened since — and a great deal has — is not in these numbers.
