The world expects AI to cut jobs, and the richer the country the deeper the fear

On 17 September 2026 the Pew Research Center published an international survey on what people expect artificial intelligence to do to work and inequality. The fieldwork, carried out between 8 February and 13 May 2026, covers 37 countries and populations: 42,151 adults interviewed in 36 countries, plus two separate surveys in the United States of 5,119 and 3,488 adults. The report is signed by Laura Silver, Richard Wike, Moira Fagan and Jordan Lippert. The main finding fits in one line: in 34 countries out of 37, the most common answer is that AI will bring fewer jobs, not more. Across the 37, the median is 46% expecting a contraction, 9% expecting growth, 25% saying they don't know.

The divide Pew puts at the centre runs along a country's income. The median expecting fewer jobs is 55% in the eighteen high-income countries and 36% in the eighteen middle-income ones; on inequality the gap repeats itself, 35% against 22% among those who think AI will widen the gulf between rich and poor. The national peaks in fear of job losses are Australia and South Korea, both at 76%, and the United States at 71%, up from 64% in August 2024. Singapore, at 46%, is the exception the report singles out. Pew places Italy among the countries that stand out for large shares of concern on both fronts, together with Australia, Canada, Greece, Poland and the United States, and notes that in Argentina, Chile, Hungary, Italy and Peru it is older people, not younger ones, who more often say AI will widen the wealth gap: an age pattern reversed compared with the rest of the sample.

About the Italian figures, it is worth saying where they come from. The country-specific percentages for Italy — 55% expecting fewer jobs, 36% expecting more inequality, 49% more concerned than excited about the everyday use of AI against 37% in the overall median, 41% reporting mixed feelings — come from the summary ANSA published on 18 September, not from the text of the Pew pages, which present those figures only in charts and appendix tables. They should therefore be credited to ANSA. One caution to add: the 55% attributed to Italy is exactly the median for the eighteen high-income countries. It may be a coincidence, but I cannot rule out that in the move from report to summary the median was read as the Italian figure: the 55% for Italy should be taken with that reservation. The sample size, too, is reported differently by secondary sources — 50,838 according to ANSA, 41,113 in 35 countries according to one aggregator — and the number to use remains Pew's.

There is a point every headline about this research risks losing: the survey measures expectations, not effects. None of the percentages concerns jobs actually created or destroyed, and the report does not separate how much of the concern comes from first-hand experience of AI and how much from media coverage of it. The findings land as Italy approaches 30 September, the date legislative decree 160/2026 aligning the country with the AI Act comes into force, whose transparency obligations have applied since 2 August 2026.

The explanation that occurs to me is that in high-income countries AI has already made its way into offices and services — but that is my hypothesis, not a finding of the survey: Pew measures expectations, not their causes. Even if it held, the interesting question is not how many jobs will disappear: it is what the people using it every day can see that the rest of us cannot yet.

— Olya