Will AI take our jobs? What the surveys actually say

Short answer

Yes, the worry is real — it is the most consistent top concern about AI across major U.S. and global surveys. But it is broader, softer, and more politically entangled than the headlines suggest. What people are expressing is anxiety about routine-cognitive work disappearing and about losing control over how AI is used on their job — not a confident mass-unemployment prediction.

Why people are concerned

Three things happened at once, and they reinforce each other.

First, the public is anxious about job loss. A Pew Research survey of more than 5,400 U.S. adults in April 2025 found 64% expect AI to lead to fewer jobs over the next 20 years, while only 5% expect more (Pew Research, April 2025). A Reuters/Ipsos poll from August 2025 reported roughly 71% of U.S. adults concerned AI will permanently displace more jobs than it creates (CNBC mirror of Reuters/Ipsos Aug 2025).

Second, workers are more worried than hopeful about AI inside their workplace. A Pew Social Trends survey from February 2025, with more than 5,200 U.S. workers, found 52% feel “worried” about AI use at work versus 36% “hopeful” (Pew Social Trends, February 2025).

Third, worry is rising alongside use. Roughly four in ten U.S. workers used AI semi-regularly or regularly by late 2025; Stanford HAI’s 2026 AI Index put global workplace AI use at 58% (Gallup Q4 2025; Stanford HAI 2026). The worry comes from the people who have touched the tools.

What is true

Several patterns hold across survey houses, countries, and years.

Jobs are the single biggest public–expert gap on AI. The same April 2025 Pew survey also asked more than 1,000 AI researchers. Seventy-three percent expected a positive impact on jobs over the next 20 years, against 23% of U.S. adults — a 50-percentage-point gap (Pew Research, April 2025). Stanford HAI’s 2026 AI Index reports the same 50-point jobs gap on its own public-opinion figure.

The anxiety is concentrated in lower-paid, service, and individual-contributor roles. Gallup’s end-of-2025 workplace poll and 2026 follow-up reporting point in this direction (Gallup Q4 2025; AP News on the Gallup poll, 2026). Pew Social Trends 2025 reinforces it: lower- and middle-income workers are more likely to expect fewer opportunities, while information/technology and banking/finance workers are the most likely to say AI will lead to more (Pew Social Trends, February 2025). A per-occupation displacement-rate table does not exist in the cited polling.

It is not a U.S.-only worry. A 25-country Pew survey in October 2025 found a 53% global median saying AI will be a “bad thing” for jobs over the next 20 years, against 19% “good thing” (Pew Global, October 2025). Stanford HAI’s 2026 index highlights India, where the share “more concerned than excited” rose 14 points year over year. The European Commission’s Spring 2025 Eurobarometer found 84% of Europeans believe AI needs careful management to protect privacy and ensure transparency in the workplace (Eurobarometer, Spring 2025; see also the global KPMG 2025 page).

The worry is widening inside the user base. Gallup’s 2026 Voices of Gen Z report found that among Gen Zers overall, the share who are “excited” about AI has dropped 14 percentage points to 22%; hopefulness fell nine points to 18%, and anger rose nine points to 31% (Gallup Voices of Gen Z 2026). Among Gen Z daily AI users, the drop in excitement is sharper — down 18 points year over year. KPMG’s 47-country longitudinal study tracks the same direction: the share saying they are worried about AI rose from 49% in 2022 to 62% in 2024, while the share saying benefits outweigh risks fell from 50% to 41% (KPMG 2025 global report PDF).

What is exaggerated, misleading, or unsupported

The “30% of jobs by 2030” framing is overstated. McKinsey’s earlier workforce-automation work estimated that up to 30% of hours currently worked across the U.S. economy could be automated by 2030 — a ceiling estimate of work hours, not a headcount of jobs lost. McKinsey’s State of AI 2025 (n=1,993 enterprise respondents) reports a much smaller workforce-impact expectation: 32% expect AI to decrease their organization’s workforce, 43% expect no change, and 13% expect increases (McKinsey State of AI 2025, via Internet Archive Wayback mirror). The “30% of jobs” figure is routinely misquoted.

Public worry is not the same as rejection. KPMG’s 2025 study reports 58% of respondents view AI systems as trustworthy. Stanford HAI’s 2026 index notes that 59% of adults globally say AI offers more benefits than drawbacks, up from 55% in 2024. People can be worried about job effects and still want the technology — they want it differently.

The partisan gap is smaller than the rhetoric suggests. Workers in information/technology and banking/finance are among the most likely to say AI use will lead to more job opportunities for them, while workers in lower- and middle-income brackets are more likely to expect fewer (Pew Social Trends, February 2025). Job-loss concern is cross-demographic rather than partisan-coded; disagreement is over causes and remedies, not whether the concern exists.

Cross-country rankings are unstable. The 53% “bad thing for jobs” figure is a median across 25 countries; per-country swings year over year can be double digits — India moved 14 points in one year. Country-rank comparisons are weaker than the cross-country consensus itself.

What remains uncertain

There is no published per-occupation displacement-rate table in the cited polling. Gallup’s Q4 2025 poll and AP News’s 2026 write-up describe the qualitative finding — exposed occupations — without a public breakdown.

There is no agreed-upon job-loss number for the next five years. McKinsey offers a 30%-of-hours ceiling. Gallup’s Q4 2025 poll found 18% of U.S. workers think AI-driven elimination in their own workplace is “very or somewhat likely” in the next five years (AP News on the Gallup poll, 2026). Those are not comparable: one is an economic model, the other is self-reported expectation.

Generation Z sentiment is one wave of year-over-year data. Whether Gen Z’s 18-point enthusiasm drop is the start of a multi-year shift or a one-cycle reaction to the 2025 deployment surge is not yet established.

Where we are likely headed

The following is editorial judgment, not a research result, on a two-to-five-year window.

The most defensible call is that job-loss worry will remain the dominant public frame on AI through 2028, even if headline job numbers do not match the worry. The pattern across Pew, Gallup, KPMG, and Eurobarometer is consistent: adoption rises, exposure rises, worry rises with both. Politicians, employers, and unions will be pressed to articulate a plan that goes beyond training; worker worry about AI in the workplace runs across income and education groups in the available data, even where it splits along partisan-coded causes rather than partisan-coded concern.

If adoption slows or a high-profile automation backlash happens, worry could plateau. If a recession or a visible mass-layoff wave tied to AI arrives, worry will spike. The most likely intermediate outcome is a slow separation between “AI worry” as a general sentiment and “AI actually cost me my job” as a smaller, sharper subset — the pattern past technology shifts have followed.

What this means for people and small businesses

For workers, the practical implication is straightforward. The cited surveys do not publish a per-occupation displacement-rate table, but Pew Social Trends 2025 finds that lower- and middle-income workers are more likely to expect fewer job opportunities, while information/technology and banking/finance workers are the most likely to say AI will lead to more. People in exposed roles do not need to panic, but they do need a plan — usually to move up the task ladder into work that requires judgment, client relationships, or physical presence, rather than work reducible to text on a screen.

For small-business owners, the surveys suggest how AI is introduced matters. Gallup’s Q4 2025 data show managers and leaders use AI more often than individual contributors, and the AP News write-up of the same poll finds workers in management, healthcare, and technology roles report more productivity benefit than workers in service roles (Gallup Q4 2025; AP News on the Gallup poll, 2026). A small rollout with an explanation is plausibly worth the time, though no survey in this brief directly measures the effect of rollout style on anxiety.

For students, the practical takeaway is that “Will AI take your job?” is the wrong question for the next few years. The better question is “How do I do work AI cannot cheaply replace?”

Bottom line

Job-loss worry is the most consistent, most cross-demographic finding in the AI survey record. It is real, it is rising alongside AI use, and it is concentrated in lower-paid and service roles. It is not a forecast of mass unemployment, and it is not the same thing as AI rejection. The watchpoints for the next two to five years are the actual displacement rate in exposed occupations, the wage trajectory there, and whether transition support can bring public sentiment back toward the optimism AI researchers currently report.