Whose Job Is Next? How Workers See AI and Job Loss

Cover Memo

This report draws on the AI@Work survey (CHIP50 Wave 38.6), a survey of 23,462 U.S. adults (weighted n = 23,485) fielded June 21–July 13, 2026. All percentages are weighted to U.S. Census population targets and are population-representative; unweighted respondent counts are reported as a reliability guide only.

These are interim estimates. The AI@Work survey launched June 21, 2026 and data collection is ongoing; the figures in this report reflect responses received through July 13, 2026 and will be revised in a later release.

The report has several bases. The job-loss items (ai_loss_current, ai_loss_ultimate), the own-job outlook item (ai_work_imp), and the openings-at-your-level item (indr_level_quantity) were asked of employed respondents / all workers (unweighted n ≈ 12,390–12,410). The entry-level hiring-change item (hiring_change) was asked only of hiring managers who report that AI has changed how they hire (unweighted n ≈ 2,557) — a small, conditional base, read as indicative. The item on whether AI advantages younger workers (ai_advantage) was asked of all adults (unweighted n ≈ 23,430). Each figure states its base. The ai_advantage scale (1 = strongly agree … 5 = strongly disagree) was confirmed against the data — agreement rises with age, the pattern one would expect if older workers, who feel disadvantaged, agree most. indr_level_quantity and hiring_change are labeled category sets (not magnitude scales) and carry no direction-reversal risk.

Key Takeaways

Introduction

The fear that AI will take people's jobs is one of the defining anxieties of the moment, and American workers feel it — but not uniformly, and not always about themselves. National surveys have found workers more worried than hopeful about AI at work: a Pew Research Center survey in late 2024 found 52% of employed adults worried about future AI use in the workplace versus 36% hopeful, and 32% expecting AI to lead to fewer job opportunities for them personally in the long run (Pew Research Center).

The AI@Work survey lets us look more closely at the shape of that fear — how it differs between the present and the future, between one's occupation and one's own job, and across the workforce. What emerges is a picture with two surprises. First, workers worry more about their line of work than about their own position: a near-majority expect AI to eventually displace people who do what they do, while only a small minority expect it to be purely bad for their own job. Second, the anxiety does not fall where the automation-fear of past decades did. It is highest not among the least-educated or manual workers, but among the young and among white-collar, data-centric workers — the people closest to what today's AI actually does.

A near-majority expect AI to take jobs like theirs — eventually

Workers see AI's threat to their occupation growing from a present minority to a near-majority in the future.

Asked whether AI is currently causing people who do work like theirs to lose their jobs, 33% of workers say yes. Asked whether it eventually will, that figure jumps to 47% — a near-majority. The 14-point gap between "now" and "eventually" is the core of workplace AI anxiety: a large share of workers who don't yet see displacement in their field nonetheless expect it to come. Only about half of workers (53%) reject the idea that AI will eventually take jobs like theirs.

Fear for the field, not for the self

Workers are far more pessimistic about their occupation than about their own job.

The same workers who worry that AI will thin the ranks of their profession are much more sanguine about their own position. Asked about AI's likely impact on their own job over the next five years, just 17% expect it to be "only negative." The rest divide between "both positive and negative" (46%), "no impact" (21%), and "only positive" (17%). So while 47% expect eventual job loss in their line of work, only about a third that many expect AI to be purely bad for them personally.

This "not my job" gap is familiar from other domains of risk perception, and it matters for how AI anxiety will play out: workers may support cushioning policies for their occupation while remaining personally optimistic, or may be underestimating their own exposure. Either way, the fear is pitched at the level of the field, not the individual.

Comparative Context

The pattern is consistent with Pew's finding that 32% of workers expect fewer personal job opportunities from AI in the long run — a personal-level figure well below the survey's 47% occupation-level expectation, exactly the field-versus-self gap this section describes (Pew Research Center).

The young fear displacement most

Younger workers — the heaviest users of AI — are also the most convinced it will take jobs like theirs.

Expectation of eventual displacement falls steadily with age. Nearly two-thirds of workers aged 18–20 (64%) and a majority of those 21–30 (52%) think AI will eventually cost people in their line of work their jobs, versus about 45% of workers in their 30s, 40s, and 50s, and just 32% of those in their 70s. The pattern holds in a regression controlling for education, income, gender, party, community type, and job type: the youngest workers have roughly twice the odds of expecting displacement as workers in their late 30s. The same gradient appears for present-day job loss (40% of the under-30s see it happening now, versus a quarter of older workers).

This is the report's central irony. The workers most fluent in AI, who use it most on the job and whom others see as advantaged by it, are also the most worried it will displace them. Both can be true at once: young workers are closer to the entry-level and routine tasks that today's AI most readily absorbs, even as they are the quickest to adopt it.

The entry-level squeeze is already visible

For younger workers, the fear is not only about the future — they are the most likely to say AI has already narrowed the openings at their level, and hiring managers report doing exactly that.

The displacement question asks workers to look ahead. A companion item asks what they see now: whether AI has changed the number of job openings at their own level. Across all workers, 16% say AI has already reduced the openings at their level, 9% say it has increased them, 24% say about the same, and the plurality — 41% — have not noticed a change (11% aren't sure). At the aggregate level, then, most workers do not yet perceive a hiring effect. But the aggregate hides a sharp age gradient that mirrors the fear almost exactly.

The workers who see the squeeze are the young. 20% of workers in their 20s and 18% of those aged 18–20 say AI has already reduced openings at their level, versus 12% of workers in their 50s, 9% of those 61–70, and just 5% of those 71–80. The mirror image is "haven't noticed a change," which climbs from 31% of workers in their 20s to 59% of those 61–70 and 67% of those in their 70s. Older, more established workers sitting at senior levels see little churn in the openings around them; younger workers, clustered at the entry level, are watching the door narrow in real time. This is the concrete, present-tense counterpart to the forward-looking fear documented above — and it falls on the same shoulders.

Hiring managers describe the same thing from the other side of the table. Among managers who say AI has changed how they hire, 27% report hiring fewer entry-level workers because of AI, and another 29% are keeping the number flat while expecting more advanced work from new hires; only 7% are hiring more (companion AI@Work hiring analysis, hiring_change; hiring managers whose hiring changed, n ≈ 2,557). Perception and practice line up: young workers feel the entry rung thinning, and the people doing the hiring say they are thinning it. That convergence is the mechanism beneath this report's central irony — the workers most fluent in AI are also the ones standing where its labor-market effect is landing first.

AI anxiety is white-collar

Unlike the automation fears of past decades, AI displacement anxiety is concentrated among knowledge workers, not manual laborers.

Whose job people think AI will take depends on the kind of work they do — and the answer inverts the familiar story. Workers whose jobs involve data processing or analysis are the most likely to expect eventual displacement (50%), while workers who do manual or physical labor are the least likely (44%). Both differences survive controls: relative to otherwise-similar workers, those in data/analytical roles have meaningfully higher odds of expecting displacement, and manual workers meaningfully lower odds. The reason is intuitive once stated — today's AI automates cognitive and desk work far more readily than physical labor, and workers appear to understand this about their own jobs.

Education, strikingly, makes no difference at all: displacement fear is essentially identical among workers with a graduate degree (46%), a bachelor's (47%), and a high-school education (47%). This sets AI anxiety apart from AI adoption, which is sharply stratified by education and income. The people most likely to use AI — the educated, the data-centric — are, if anything, more likely to fear it, not less. AI displacement anxiety is broadly democratic across the workforce, tilted toward the young and the white-collar rather than toward the disadvantaged.

Who holds the edge: the younger-worker advantage

A plurality of Americans think AI advantages younger workers — and older workers, who feel it most, agree the most.

Beyond who fears losing work, there is the question of who AI helps and hurts within a workplace. Asked whether "AI gives younger workers an advantage over older or longer-tenured workers," 49% of adults agree (17% strongly), 32% are neutral, and 20% disagree. Agreement rises steadily with age: older adults — the ones who would be on the losing end of such an advantage — are the most likely to endorse the statement, while the youngest are closer to neutral.

Set against the displacement findings, this completes a nuanced picture of generational AI politics. Older workers believe AI tilts the workplace toward the young; younger workers, for their part, are the most afraid of being displaced by it. These are not contradictory so much as two sides of a churning entry-level labor market — one where AI fluency is a young worker's asset and the automatable, entry-level task is a young worker's exposure.

Conclusion

American workers are not panicking about AI, but they are bracing for it. A near-majority expect it to eventually cost people in their line of work their jobs, even as most remain personally optimistic about their own position — a field-versus-self gap that will shape how the politics of AI and work unfold. The anxiety is broadly shared, but where it concentrates upends the old automation narrative: it is highest among the young, who use AI most, and among white-collar, data-centric workers, whose cognitive tasks are exactly what today's systems do best. Education, the great divider of AI use, does nothing to divide AI fear.

That combination — the people closest to AI being the most worried about it — is the finding to carry forward, and it is no longer only a matter of expectation. Younger workers are already the most likely to say the openings at their level have thinned, and the managers who hire them report cutting entry-level roles for exactly that reason; perception and practice have begun to meet. As the data collection for this wave continues and the technology advances, the workers to watch are not only the manual laborers of the automation imagination but the young analysts, writers, and coders who have adopted AI fastest and expect it, sooner than anyone, to come for jobs like theirs.

Appendix A — Methods

Data source. AI@Work survey — CHIP50 Wave 38.6 (Civic Health and Institutions Project / COVID States Project). Total unweighted n = 23,462; weighted n = 23,485. Fielded June 21–July 13, 2026. Data collection was ongoing at the time of this extract — figures reflect responses through July 13, 2026 and are preliminary. Estimates are weighted to U.S. Census population targets.

Measures and bases. ai_loss_current and ai_loss_ultimate ("Do you think AI is currently / will eventually cause people who do work like yours to lose their jobs?"; Yes/No) and ai_work_imp (impact of AI on one's own job in the next five years; only positive / both / only negative / no impact) were asked of employed respondents (unweighted n ≈ 12,390–12,400). indr_level_quantity ("Has AI affected the number of job openings at your level?"; fewer / about the same / more / not noticed a change / not sure) was asked of all workers (unweighted n ≈ 12,413). hiring_change (entry-level hiring strategy under AI) was asked only of hiring managers who report that AI changed how they hire (unweighted n ≈ 2,557). ai_advantage ("AI gives younger workers an advantage over older or longer-tenured workers"; 5-point agree–disagree) was asked of all adults (unweighted n ≈ 23,430).

Coding. ai_advantage is coded 1 = strongly agree … 5 = strongly disagree, confirmed against the data (the weighted mean falls monotonically with age — older respondents agree more, coherent only under this direction). indr_level_quantity (1 = fewer, 2 = about the same, 3 = more, 4 = not noticed a change, 5 = not sure) and hiring_change (1 = fewer entry-level, 2 = same #/same expectations, 3 = same #/more advanced work, 4 = more entry-level) are labeled category sets reported per the authoritative codebook; they are nominal, not magnitude scales, so they carry no direction-reversal risk and are reported categorically. Two binary indicators were derived for analysis: loss_current_bin and loss_ultimate_bin (1 = Yes).

Estimation. Cross-tabulations are weighted shares. One weighted logistic regression (survey-weighted GLM, logit; n = 12,396; McFadden pseudo-R² = 0.016) models expecting eventual displacement on age, education, gender, income, party, community type, and two job-type indicators (manual labor; data/analysis). The very low pseudo-R² is itself a finding: displacement fear is broadly shared rather than concentrated in any one group. A difference is called independent only where the coefficient is significant at p <.05; raw crosstab gaps are descriptive. Cells with unweighted n < 10 are suppressed or flagged (workers 80+, unweighted n = 21).

Reproducibility. Figures map to Appendix B tables; values copied verbatim from the underlying tabulations. Wording and coding sourced from the Qualtrics instrument via the project's AI@Work survey verified-updates crosswalk.

Appendix B — Data tables

Agreement rises with age: the weighted mean moves from 3.08 (near neutral) among workers 18–20 to about 2.5 (leaning agree) among workers 60 and older.

Appendix C — Question wording (verbatim)

• ai_loss_current — "Do you think that AI is currently causing people who do work like yours to lose their jobs?" (Yes; No)
• ai_loss_ultimate — "Do you think AI will eventually cause people who do work like yours to lose their jobs?" (Yes; No)
• ai_work_imp — "In your opinion, what would be the impact of AI tools like ChatGPT on your own job in the next 5 years?" (Only positive impact; Both positive and negative impact; Only negative impact; No impact)
• ai_advantage — "How much do you agree or disagree with the following statement: AI gives younger workers an advantage over older or longer-tenured workers." (Strongly agree; Somewhat agree; Neither agree nor disagree; Somewhat disagree; Strongly disagree)
• indr_level_quantity (all workers) — "Has AI affected the number of job openings at your level?" (Fewer; About the same; More; Have not noticed a change; Not sure)
• hiring_change (hiring managers whose hiring changed) — entry-level hiring strategy under AI (Fewer entry-level roles because of AI; Same number, same expectations; Same number, but expect more/higher-level work with AI; More entry-level roles because of AI)