AI job loss is now the majority expectation in almost every country on earth. Pew Research Center published a global survey on 17 September 2026 covering 42,151 people across 37 countries, and in 34 of those 37 countries people are more likely to believe AI will lead to fewer jobs over the next 20 years than to more of them. That is not a plurality in a handful of anxious economies. It is close to unanimity across the sample.

The fear is strongest where the technology is most visible. Worries about AI job loss run particularly high in Australia at 76 percent, South Korea at 76 percent and the United States at 71 percent — three wealthy, highly connected economies with large white-collar workforces. The American figure has risen seven percentage points since Pew asked the same question in 2024.

There is one detail that shapes how the whole survey should be read, and The Verge’s Thomas Ricker put it plainly: the fieldwork ran from 8 February to 13 May 2026, well ahead of the recent wave of apocalyptic warnings from AI executives and researchers. Everything measured here predates the September arguments about pacing the frontier, embedded evaluators and industry self-regulation. Whatever these numbers are, they are not a reaction to the last month of headlines. This article sets out what Pew asked, where AI job loss fear concentrates, which demographic splits are real, what the survey says about inequality, and what an employer should sensibly do with the finding.

What the Pew AI Job Loss Survey Actually Measured

ai job loss pew global survey 37 countries b hard hat dome with a short front brim

Precision matters here, because “people are worried about AI” is a headline that survives almost any data. The specific questions are narrower and more useful.

The sample

42,151 respondents across 37 countries, with the United States surveyed separately and in larger numbers alongside the international sample. Fieldwork ran 8 February to 13 May 2026.

The countries covered

Argentina, Australia, Bangladesh, Brazil, Canada, Chile, Colombia, France, Germany, Ghana, Greece, Hungary, India, Indonesia, Israel, Italy, Japan, Kenya, Malaysia, Mexico, the Netherlands, Nigeria, Pakistan, Peru, the Philippines, Poland, Singapore, South Africa, South Korea, Spain, Sri Lanka, Sweden, Thailand, Turkey, the UK, the US and the West Bank and East Jerusalem. That spread is what makes the AI job loss result a global finding rather than a Western one.

The 20-year framing

Respondents were asked about the effect on jobs over the next 20 years, not next year. Long-horizon questions invite speculation, which is exactly why the near-unanimity across 34 of 37 countries is notable: people are not being asked to forecast a quarter.

The uncertainty bucket

Around 20 percent of respondents in many countries said they did not know, and in some middle-income countries more than 40 percent did. AI job loss expectations are not evenly informed across the sample, and the uncertain share is largest where AI deployment is least visible.

What it does not measure

The survey records expectations, not outcomes. Nobody in it has counted a redundancy. It is a map of what populations believe about AI job loss, which is a real and consequential thing, but it is not labour-market data.

CountryExpect AI to reduce jobsNotable context
Australia76%Joint highest in the survey
South Korea76%Yet 61% say equally concerned and excited
United States71%Up 7 points since 2024
34 of 37 countriesMajority or pluralityJob losses outweigh job creation
Several middle-income countriesLower net figureDriven by 40%+ answering “don’t know”

Where AI Job Loss Fear Runs Highest, and Why

ai job loss pew global survey 37 countries c punch clock block with one round blank dial

The geography of this result is the opposite of what a naive model would predict, and the pattern repeats across several of Pew’s questions.

Wealth correlates with worry

The countries most alarmed about AI job loss are among the richest and most technologically saturated. Australia, South Korea and the United States sit at the top; several lower-income economies sit lower, with much larger uncertain shares.

Exposure, not vulnerability

One reading is that fear tracks exposure to AI products rather than vulnerability to automation. In economies where AI tools are already in daily use at work, people have a concrete mental picture of what a system might take over.

The white-collar concentration

Anthropic chief executive Dario Amodei’s warning that AI could eliminate half of entry-level white-collar jobs is the specific claim most often cited in coverage of AI job loss fear. Rich economies have the largest white-collar sectors, so that warning lands hardest there. The tasks named as most exposed — drafting, summarising, triaging correspondence — are precisely the ones natural language processing has become competent at in the past two years.

Where the informed are more worried

Separate research this year found that people who understand AI better tend to be more worried about losing their jobs to it, not less — which cuts against the standard assumption that fear is a product of unfamiliarity.

Where uncertainty dominates instead

In several middle-income countries the largest single answer is neither optimism nor AI job loss fear but “don’t know.” That is not complacency; it is a measure of how unevenly the technology has actually arrived.

AI Job Loss Is Not the Only Thing People Expect to Get Worse

ai job loss pew global survey 37 countries d three descending solid blocks in a row tall to short

The second half of Pew’s finding is about distribution rather than employment, and the two reinforce each other.

The inequality result

Across the survey, a median of respondents expect AI to increase the gap between rich and poor rather than narrow it. This is a companion fear to AI job loss and, in several countries, a stronger one.

The countries most convinced

South Korea, Australia, the United States, Greece and Canada all record 40 percent or more expecting AI to widen inequality — largely the same list that leads on AI job loss.

Almost nobody expects improvement

No country in the survey exceeds 25 percent believing AI will decrease inequality, and in 29 of the 37 countries that figure is in single digits. That is the most one-sided number Pew reports.

The ideological split

In 8 of the 28 countries where the question was asked, left-leaning respondents were more likely than right-leaning ones to cite inequality concerns. The United States is the extreme case: liberals were 31 points more likely than conservatives to name it.

Why the two fears compound

An economy can absorb AI job loss if the gains are broadly shared, and it can absorb rising inequality if employment holds. The survey finds populations expecting both at once, which is the combination with the least political slack in it.

Share expecting AI to reduce the number of jobs, selected countries
Australia 76%
South Korea 76%
United States, 2026 71%
United States, 2024 64%
US adults aged 18 to 34, 2026 55%

Who Fears AI Job Loss Most Within Each Country

ai job loss pew global survey 37 countries e piggy bank with a rounded body and one narrow top slot

The demographic splits are where the survey stops being a headline and starts being useful.

Young adults are moving fastest

Among Americans aged 18 to 34, the share expecting AI job loss rose from 40 percent in 2024 to 55 percent in 2026 — a 15-point move in two years. Pew records sharp increases in concern among younger adults in Sweden, Poland, Japan, Australia, Brazil and the US.

Young adults are also the most ambivalent

The same cohort is most likely to report mixed feelings overall. They are not uniformly hostile to AI; they are simultaneously the most engaged with it and the most anxious about what it does to the labour market they are entering.

Older adults worry about a different thing

Adults aged 50 and over are more concerned than younger people about AI’s effect on daily life in 15 countries. The generational split is not “who is worried” but “worried about what” — daily life for the older group, AI job loss for the younger.

The gender gap is real but modest

In 11 mostly high-income countries women were more likely than men to expect job losses. France shows the widest split at 10 points.

Education and income cut differently by development level

In middle-income countries, higher-income and better-educated respondents were more likely to predict AI job loss, while lower-income groups were more likely to answer “don’t know.” The people closest to the formal economy are the people expecting the disruption.

SplitDirectionWhere it shows up
Age 18–34More likely to expect job lossesCanada, France, Singapore, Sweden, India, Indonesia, Malaysia
Age 50+More concerned about AI in daily life15 countries
Women vs menWomen more likely to expect job losses11 mostly high-income countries; France +10 points
Higher income and educationMore likely to expect job lossesMiddle-income countries
Left-leaningMore likely to cite inequality8 of 28 countries; US gap 31 points

The Number That Complicates the AI Job Loss Headline

ai job loss pew global survey 37 countries f milestone marker stone with a rounded top

Pew’s own data contains a strong counterweight to the “destroyer of jobs” framing, and it deserves equal billing.

Ambivalence beats alarm

A global median of 41 percent say they feel equally concerned and excited about AI’s growing presence in daily life. Only 37 percent are primarily concerned, and 13 percent are primarily excited. The largest single group is not frightened — it is torn.

The South Korea paradox

South Korea records one of the highest AI job loss expectations in the survey at 76 percent, and simultaneously reports 61 percent saying they are equally concerned and excited. A population can expect disruption and still want the technology.

The one net-positive country

Israel is the only country in the survey where excitement exceeds concern. Every other national sample leans the other way or splits evenly.

What this does to the narrative

“AI is feared globally as the destroyer of jobs” is an accurate reading of the employment question. It is an incomplete reading of the survey, because the same people register genuine enthusiasm about AI in daily life at close to the same rate.

The useful synthesis

The public position that best fits the data is conditional acceptance: broadly willing to use AI, broadly expecting AI job loss, and broadly expecting the benefits to be distributed badly. That is a demand for governance rather than a demand for prohibition.

What the Timing of the Survey Leaves Out

The fieldwork window is the single most important caveat on the whole dataset.

Three months, ending in May

Responses were collected between 8 February and 13 May 2026. Anything that happened after mid-May is invisible to these numbers.

What happened after

The summer and early autumn produced the Hugging Face agent incident, a sequence of executive essays calling for slowdowns or regulation, an open letter from prominent researchers, and a public dispute between major labs about AI safety. None of it is reflected here.

Which direction the bias runs

It is reasonable to expect the next wave to show higher concern, not lower, given what has been in the news since. The 71 percent US figure is therefore best read as a floor rather than a peak.

The comparison point that survives

Because Pew asked the same question in 2024, the seven-point US rise and the 15-point rise among under-35s are measured on consistent wording. Trend data with stable wording is more durable than any single-year level.

Why it still matters

A survey taken before the panic tells you the AI job loss expectation is structural rather than a news-cycle artefact. That is a stronger finding than one taken last week would have been.

How AI Job Loss Fear Compares With What Employers Report

Expectations and measurements are different objects, and 2026 produced both. Setting them side by side is the fastest way to see how much of the AI job loss picture is anticipation.

The employer-side number

Mercer reported that employee concern about job loss due to AI rose from 28 percent in 2024 to 40 percent in 2026 — a different instrument, a different population, and the same direction of travel as Pew’s.

The worker-side number in the US

A 2026 Resume Now survey found 60 percent of American workers believe AI will eliminate more jobs than it creates over the coming year. That is a one-year horizon rather than Pew’s twenty, and it still lands in majority territory.

The forecaster-side caution

Forrester’s position is that AI-led disruption will escalate while fears of a job apocalypse are overstated, and that over half of layoffs attributed to AI will be quietly reversed as companies meet the operational reality. That is the strongest available counterweight to the AI job loss consensus.

Why the two can both be right

Anticipated displacement and realised displacement diverge when the technology is easier to announce than to deploy. A company can credibly plan for AI job loss, communicate it, and then discover the workflow will not survive without the person it removed.

What that means for the survey

Pew is measuring belief, and belief moves faster than payroll. The seven-point US rise since 2024 is a real change in what people expect, and it is not evidence that seven percent more jobs have gone.

SourceWhat it measuredHeadline figure
Pew Research CenterExpected effect on jobs over 20 years, 37 countries34 of 37 expect fewer jobs
MercerEmployee concern about losing a job to AI28% in 2024 to 40% in 2026
Resume NowUS workers, one-year outlook60% expect net job destruction
ForresterForecast of realised disruptionEscalating, but apocalypse overstated

How to Read a Global AI Job Loss Survey Without Being Misled

Cross-country polling invites four specific mistakes. Each of them shows up in coverage of this release.

Do not compare levels across countries casually

Question wording is translated, and willingness to express pessimism to a pollster varies culturally. The within-country trend against 2024 is far more robust than any league table of AI job loss percentages.

Watch the denominator

A country where 45 percent expect job losses and 40 percent say “don’t know” is not more optimistic than one at 70 percent. It is less informed. Several middle-income results in this survey have exactly that shape.

Do not treat a 20-year question as a forecast

Twenty years is beyond the planning horizon of almost every respondent. What the AI job loss question really captures is a disposition towards the technology, expressed through the most concrete channel available.

Separate the two questions Pew asked

Employment and inequality are distinct items with distinct national profiles. Merging them into “people hate AI” throws away the finding that a plurality feel equally concerned and excited.

Check the fieldwork window before quoting

This one closed on 13 May 2026. Any claim that the survey shows a reaction to the autumn AI safety debate is simply false, and the dates are published.

What Employers Should Take From Global AI Job Loss Fear

The practical question for anyone running a team is what to do with a workforce that mostly expects this technology to cost jobs.

Assume the expectation is already in the room

Seventy-one percent in the US and 76 percent in Australia means most of your staff hold this view whether or not they have said so. A rollout plan that does not address AI job loss expectations explicitly is addressing them implicitly, badly.

Separate augmentation from headcount

The distinction between redesigning work and reducing headcount is the one employees actually care about. Saying which one you are doing, in specific terms, is worth more than any reassurance about the technology itself.

The reversal risk is real

Analysts have noted that a meaningful share of layoffs attributed to AI get quietly reversed when the operational reality lands. Announcing AI job loss you cannot deliver on damages trust twice — once at the announcement and again at the rehire.

Young staff are the most exposed group

The steepest rise in AI job loss fear is among 18-to-34-year-olds, and entry-level white-collar work is the category most often named as at risk. That is also your pipeline.

Governance is the credible answer

The survey’s inequality finding suggests people want the gains shared, not the technology banned. Concrete internal commitments — retraining budgets, redeployment guarantees, a published policy — answer the actual concern. Organisations working through an AI readiness assessment should treat workforce communication as a deliverable rather than an afterthought, and pair it with an honest view of where machine learning genuinely changes a role.

AI Job Loss: Frequently Asked Questions

What did the Pew survey find?

That in 34 of 37 countries surveyed, people are more likely to believe AI will lead to fewer jobs over the next 20 years than to more jobs. The survey covered 42,151 respondents.

Which countries are most worried?

Australia and South Korea at 76 percent, and the United States at 71 percent. All three are wealthy economies with large white-collar workforces.

When was the survey conducted?

Between 8 February and 13 May 2026 — before the wave of AI safety warnings and executive essays that dominated coverage in August and September.

Has US concern about AI job loss increased?

Yes. The US figure rose seven percentage points since Pew asked the same question in 2024. Among Americans aged 18 to 34 it rose from 40 percent to 55 percent.

Do people also expect AI to worsen inequality?

Yes. A median expects AI to widen the gap between rich and poor, no country exceeds 25 percent expecting it to narrow, and in 29 countries that figure is in single digits.

Is the overall mood negative?

Not entirely. A global median of 41 percent feel equally concerned and excited, against 37 percent primarily concerned and 13 percent primarily excited. Israel is the only country where excitement exceeds concern.

What specific claim drives the coverage?

Anthropic chief executive Dario Amodei’s warning that AI could eliminate half of entry-level white-collar jobs is the most frequently cited forecast in reporting on AI job loss.

References