Pew Research's new 37-country survey finds that in 34 of them, people expect AI to destroy more jobs than it creates over the next 20 years. The poll of 42,151 respondents, fielded between February 8 and May 13, 2026, lands as executives at frontier labs escalate their own warnings about white-collar displacement. Job-loss fears run highest in wealthy economies: 76% in Australia, 76% in South Korea, and 71% in the United States.
The three-country gap — 34 out of 37 — is the headline number, because it collapses the usual regional and political splits. Wealthier and poorer nations, democracies and autocracies, tech-exporting and tech-importing economies all converge on the same expectation. This is not a Western anxiety or a developing-market anxiety. It is a global one.
Sentiment, however, is not uniformly bleak. A global median of 41% say they feel equally concerned and excited about AI's growing presence in daily life, against 37% who are primarily concerned and 13% primarily excited. That leaves a plurality who see the technology as a mixed bag rather than a threat — a nuance often lost in the headline framing.
Key facts
- 01Pew Research surveyed 42,151 people across 37 countries between February 8 and May 13, 2026.
- 02In 34 of the 37 countries, majorities expect AI to cause net job losses over the next 20 years.
- 03Job-loss fears peak in Australia (76%), South Korea (76%), and the US (71%).
- 04Globally, 37% are primarily concerned about AI, 13% primarily excited, and 41% feel both equally.
- 05The poll closed before Anthropic CEO Dario Amodei warned AI could eliminate half of entry-level white-collar jobs.
Age matters. Older respondents are more likely to worry about AI's day-to-day impact, while adults aged 18 to 34 skew toward mixed feelings. But the younger cohort is also where anxiety is climbing fastest. In Sweden, Poland, Japan, Australia, Brazil, and the US, concern among under-35s has risen sharply over the past year, closing the gap with older generations who were already worried.
Respondents also expect AI to widen income inequality, with the strongest agreement in wealthy left-leaning countries. That framing matters for the policy conversation: if the public already assumes AI is a concentration technology, the political appetite for redistribution, retraining subsidies, and antitrust scrutiny grows before any specific layoff cycle hits.
The survey closed in mid-May, which means it does not capture the reaction to Anthropic CEO Dario Amodei's warning that AI could eliminate half of entry-level white-collar jobs. Amodei's comments, made publicly in the months after fieldwork ended, gave the job-loss thesis its most prominent industry endorsement to date and are likely to have hardened the numbers Pew reports.
Pew's framing — job losses over 20 years — is a long horizon, which cuts both ways. Two decades is enough time for entirely new categories of work to emerge, as they did after prior automation waves. It is also long enough that expectations of decline can become self-fulfilling if firms front-load hiring freezes and students steer away from fields they believe will be automated.
The specific country pattern is worth noting. High-fear economies like Australia, South Korea, and the US share a large share of employment in exactly the office-based, information-processing roles that current models handle best. In those labor markets, the abstract 20-year question is closer to a concrete near-term forecast for a meaningful slice of the workforce.
The counterweight in the data is the excitement-and-concern group. A 41% plurality that holds both views simultaneously is not a public prepared to reject AI outright — it is a public that expects to negotiate the terms. That distinction will shape whether the next few years of AI policy focus on adoption incentives, worker protections, or both.
The public has already priced in a version of Amodei's forecast, and it did so before he made it. That reframes the political and commercial task for AI companies. The question is no longer whether to convince the public that disruption is coming — that argument is over. The question is whether the labs, the customers deploying their models, and the governments regulating them can point to any concrete mechanism by which the jobs displaced are replaced by better ones. Absent that, the 34-of-37 number is the ceiling of public patience, not the floor.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
