In 34 of 37 Countries, People Expect AI to Cost Jobs

In 34 of the 37 countries Pew Research Center covered in a report published on 17 September 2026, more people expect artificial intelligence to reduce the number of jobs in their country than to increase it. The fieldwork ran from 8 February to 13 May 2026 and reached 42,151 adults across 36 countries. The United States is the thirty-seventh, surveyed separately through the Center's American Trends Panel: 5,119 adults from 17 to 23 February 2026 and 3,488 from 22 to 28 June 2026.

Worry rises with GDP per capita

The clearest pattern in the data is an income gradient, and it runs the opposite way to the one a reader might expect. Anxiety about AI and employment is highest in the wealthiest countries surveyed. In Australia, South Korea and the United States, around seven in ten adults or more say AI will result in job loss over the next 20 years. In the United States, the only country where Pew has asked the question before, that share has risen 7 percentage points in two years.

In many middle-income countries, by contrast, the dominant answer is that people do not know. About a fifth of adults or more are unsure how AI will affect jobs in their country, and in some middle-income nations the unsure share reaches around four in ten or more. Pew notes that lower-income and less-educated respondents in those countries are the most likely to give that answer, which means cross-country rankings in the middle of the table should be read loosely.

Almost nobody thinks it narrows the gap

On inequality the balance of opinion is one-sided. More people say AI will widen the gap between rich and poor than say it will narrow it, and in South Korea, Australia, the United States, Greece and Canada around four in ten adults or more expect it to widen. There is no country in the survey where more than one in four think AI will reduce economic inequality, and in 29 of the 37 that figure is in single digits.

Within countries, the splits are political and generational. American liberals are 31 percentage points more likely than conservatives to say AI will increase inequality, one of the largest ideological gaps Pew recorded. Americans under 35 are well ahead of those aged 50 and over on the same question, at 56% against 38%.

Concern is moving, and the young are moving fastest

Five countries show a measurable rise since the 2025 survey in the share who are more concerned than excited.

The sharpest movement is by age. Among Americans aged 18 to 34, the share who are primarily concerned about AI has gone from 40% in 2024 to 55% in 2026. They are now as likely as those aged 50 and over to say they are mainly concerned, reversing the pattern of five years ago. In Sweden the age gap has all but closed over a single year. Concern also grew among over-50s in France, Hungary and the Netherlands.

Concern is not the same as rejection. Across the 37 countries, a 37% median say they are more concerned than excited about the increased use of AI in daily life, while a larger 41% median say they are equally concerned and excited, which is the most common answer in about half the countries. Israel is the only nation where more people describe themselves as primarily excited than primarily concerned. South Korea makes the point sharply: South Koreans are among the most likely anywhere to forecast job losses and widening inequality, yet 61% say they are equally concerned and excited and 21% say they are primarily excited.

Knowing more does not mean worrying more

The finding that cuts against the grain concerns awareness. In around half the countries surveyed, people who have heard or read a lot about AI are more likely than the less aware to say it will cost jobs. In many of the same countries, those high-awareness respondents are also more positive about the technology overall. Germany shows both at once: among Germans who have heard a lot about AI, 26% are primarily concerned, 59% are equally concerned and excited and 15% are primarily excited, against 41%, 51% and 8% among those who have heard little or nothing.

The honest reading is that information is not pushing opinion in a single direction. It appears to make people more specific rather than more frightened: more willing to name job losses as the likely cost, and less willing to describe the whole technology as something to dread. Whether that holds as agents move into more workplaces is exactly what the next wave will test.

Two limits are worth stating. Most of the fieldwork closed on 13 May 2026, before the summer's disclosures about autonomous agents and before this autumn's fight in Washington over who pays for data-centre power, so none of that is priced into these answers. And a survey measures expectations, not payrolls. The report's own companion chapter asks a question with more immediate policy weight, which is whether people trust China, the United States or the European Union to regulate AI at all.