Global Survey: Most Countries Expect AI to Bring Job Loss
More people in 34 of 37 countries expect artificial intelligence to cause net job loss over the next 20 years rather than create new jobs, according to Pew Research Center survey data published September 17, 2026.
The report, titled "Globally, More People Expect AI to Cause Job Loss Than Growth," draws on interviews with 42,151 adults across 37 countries conducted from February 8 to May 13. Pew Research Center The sample is large for AI attitude research. The gap between loss and gain is wide.
In only three surveyed countries did expectations not tilt toward displacement. Elsewhere, expecting loss was the most common answer. Concern was most pronounced in Australia, where 76% said AI will lead to job losses, in South Korea, also at 76%, and in the United States, at 71%. The Verge USA Today reported the same central finding.
Respondents were also more likely to say AI will increase the gap between rich and poor. That view held globally, not only in high-income economies where business use of AI models, copilots (assistants built into software) and automation tools is furthest advanced.
Attitudes toward daily use were less one-sided than the employment question. A global median of 41% said they feel equally concerned and excited about AI's growing presence in daily life. A median of 37% said they feel mainly concerned, compared with 13% who said they feel mainly excited.
Age remains a dividing line. Older people were more likely than younger people to report concern about AI in daily life. At the same time, concern among younger adults about AI-driven job loss and inequality rose sharply over the past year in Sweden, Poland, Japan, Australia, Brazil and the United States.
Earlier Pew work provides context for how those views are distributed. Pew found a strong link between a country's income as measured by GDP per capita and awareness of AI, in research published October 15, 2025. Pew Research Center In a separate 25-country measure, a median of 37% of adults said they trust the United States to regulate AI effectively, while a median of 48% said they do not. In the U.S. itself, half of adults said in a June 2025 survey that increased use of AI in daily life makes them feel more concerned than excited.
The broader context here is worth spelling out. Public expectations of labor market effects often run ahead of measured adoption by companies, and they combine distinct mechanisms: task automation, hiring freezes, outsourcing of entry-level knowledge work, algorithmic management (software used to set schedules and assess performance), and wage effects when productivity gains accrue unevenly. The survey does not separate those channels. It captures a mood about direction, including exposure without a settled verdict, not a forecast of size or timing.
In my view, that mood deserves to be taken seriously without being read as a technical verdict. I have watched my own children move from skepticism to casual reliance on AI assistants in school and early work, a shift that took months rather than years. Familiarity did not erase concern about cheating, deskilling or entry-level hiring. It made those concerns more concrete. For technology professionals, the signal in this data is less about model capability than about legitimacy. Systems that change hiring, scheduling, evaluation and access to expertise will be judged on employment and distributional effects, alongside accuracy, inference latency (the delay before a model responds) and safety. The long arc still favors tools that expand what small teams and individuals can do, but that outcome will depend on deployment choices that give workers observable leverage, not only employers observable efficiency.
