Pew asked 42,000 people in 37 countries whether AI will cost jobs. In 34 of them, more said yes than no

A Pew Research Center survey of 42,151 adults across 36 countries, fielded February through May 2026, finds that in 34 of 37 places surveyed, people are more likely to expect AI will mean fewer jobs over the next 20 years than more. The count reaches 37 because Pew ran two separate US surveys alongside the 36-country wave -- one of 5,119 adults in February, one of 3,488 in June -- both asking a version of the same question. The pattern holds across income levels, but not evenly: a median of 55% of adults across 18 high-income countries expect fewer jobs, against 36% across 18 middle-income countries.

In the United States specifically, 71% of adults now say AI will lead to fewer jobs over the next two decades, up from 64% in 2024 -- a 7-point rise in two years. The increase was not even across age groups. Adults under 30 moved from 61% to 73% over the same period, a 12-point jump, nearly double the rate of the population as a whole. Pew's own short-read on the finding, published in August, frames it plainly: younger US adults, who might be expected to be the most comfortable with new technology, are now the age group most convinced it will cost jobs rather than create them.

The age split sharpens the same story the country-level numbers already tell: the people with the most working years still ahead of them are the ones least confident those years hold as many jobs as they expected.

Wealth correlates with concern here more than with most other AI attitudes Pew tracks. Job-loss expectations track income level closely -- that 55%-versus-36% gap -- but general excitement-versus-concern about AI in daily life doesn't track nearly as tightly with a country's GDP per capita, according to Pew's companion report on the same fieldwork. Kenya, Nigeria and the Philippines register the lowest job-loss concern among the countries surveyed; Australia and the US sit among the highest. Uncertainty itself is unevenly distributed too -- in several middle-income countries, roughly two-fifths of adults or more say they simply don't know which way AI will move employment, versus a smaller, but still real, share in wealthier countries.

The same fieldwork asked a second question that gets less attention than the jobs number but points at a sharper worry: whether AI will widen the gap between rich and poor. There, the income split is even starker -- a median of 35% of adults in high-income countries say AI will worsen inequality, against 22% in middle-income countries, and the country-level spread is wide even within the wealthy group: 46% of US adults say so, versus 13% in Colombia. Job loss and inequality are related worries but not the same one -- a country can expect AI to shrink the total number of jobs while still disagreeing about whether the jobs that remain will be distributed any more fairly than they are now, and Pew's data suggests richer countries are more pessimistic on both counts at once.

None of this measures what AI is actually doing to employment right now -- it measures what people *expect* it will do over the next twenty years, which is a different question. A July analysis of 2026 layoff-attribution data found that question genuinely hard to answer even with hard numbers in hand: companies citing "AI" for a layoff are making a communications choice as much as a technical one, and no consensus exists on what share of this year's actual job losses are AI-driven versus ordinary economic conditions. Pew's survey captures a belief, not a measurement -- and the gap between the two is exactly what makes the belief itself newsworthy. Whether or not the job losses arrive on the schedule people expect, the expectation is already something people are planning around.

The uncertainty share is itself worth sitting with. In a decision this consequential, roughly a fifth of adults globally -- more in poorer countries -- say they don't know which way it goes, not that they're unconcerned. That's not indecision so much as an honest accounting of how little firm evidence either side of the debate actually has to offer a worker trying to plan the next twenty years of a career.

Roughly a fifth of adults, more in poorer countries, say they simply don't know which way AI moves the number of jobs -- not that they're unconcerned, but that nobody has given them reason to be confident either way.