AI Job Losses Rare as Only 3% of Workers Report Displacement

A survey that undercuts the panic, and the hype

Nearly four years after ChatGPT launched, a large survey of American workers suggests that, by their own account, AI has done much less to their careers than either the doomsayers or the boosters predicted. Jeffrey C. Dixon, a sociologist at the College of the Holy Cross, commissioned YouGov to poll 1,250 employed U.S. adults between July 30 and Aug. 4, 2026, about how they use AI, what they expect from it, and how it has already affected their job prospects.1 Dixon calls himself an AI pessimist, and he said he was surprised by how few respondents reported losing work because of the technology.1

The headline numbers are small. About 3% of workers said they had lost a job to AI since 2023. Roughly 6% said they had moved into a job that did not exist before AI adoption, such as data labeling or AI compliance work, and about 9% linked a promotion or advancement to AI.26 Put the other way, roughly 95% said no to the job-loss question, 90% said no to the new-job question and 88% said no to the promotion question. Another 3% to 4% weren't sure.13

The study is still in progress and has not been peer-reviewed, a caveat that some republished versions dropped.3 Dixon's funding came from a College of the Holy Cross grant and a provost's office account, and he also runs an AI ethics consulting business.4

The most interesting result is promotion, not loss

The public debate frames AI almost entirely as a question of who gets fired. Dixon's data points somewhere else. The biggest effect workers reported was a chance to move up, not a pink slip.35 His reading is that AI skills may be helping some workers climb, but it is too early to know whether that edge will last. Agentic AI, which can act on its own and automate more of a job, could make today's AI skills less valuable.5

Other research backs up this upside. PwC's 2026 Global AI Jobs Barometer, built on more than a billion job ads across 27 countries, found that roles requiring AI skills pay an average of 62% more than comparable roles without them, up from 57% a year earlier.17 Those surveys measure different things, so they don't confirm each other directly. Still, both suggest that workers who can use the tools well are capturing real gains while losses stay limited for now.

Why a small number can still matter

Dixon is careful about what his numbers can and cannot show, and his caveats deserve weight. Job changes are uncommon in any given period, so a few percent reporting AI-related changes has to be read against that low baseline.1 Workers may also not know whether AI actually caused a change in their job.1 The biggest limitation is that everyone surveyed had a job. People pushed out of work, by AI or anything else, were not in the sample.13 The YouGov panel is also opt-in, so its answers may differ from those of randomly selected workers.1

That gap matters, because the clearest AI effects show up among people trying to get into the labor market, not people already in it. An incumbent-worker survey will mostly miss that group by design.

Where the wider evidence agrees

On the aggregate picture, Dixon's findings match a growing body of research. The Yale Budget Lab, using Current Population Survey data through the first quarter of 2026, found no clear economy-wide effect on employment or wages.11 Anthropic's March 2026 labor market report likewise found no clear jump in unemployment among workers in the most AI-exposed occupations.13 In Denmark, Humlum and Vestergaard linked adoption surveys to administrative records for 25,000 workers. Even where 93% of workers in firms with enterprise chatbots said they used AI, they found no effect on earnings, hours or wages.9

Executives tell a similar story. In a survey of nearly 6,000 executives in the U.S., U.K., Germany and Australia, 69% of firms said they use AI, but about 90% said it has not meaningfully changed overall employment or productivity yet.9 Goldman Sachs economists describe AI's labor market impact as "visible but narrow," with drags in marketing, graphic design, customer service and some tech jobs. They note those drags are partly offset by about 209,000 extra construction jobs tied to data centers since 2022.14

Where it diverges: the entry-level squeeze

The broad calm hides a more troubling trend among young workers, and on this point the research mostly agrees. Stanford researchers using ADP payroll data found that employment of 22- to 25-year-olds in highly AI-exposed jobs was 19% below where it would have been had it kept pace with less-exposed peers by June 2026, up from 15% a year earlier.9 The researchers say the gap comes mostly from fewer hires rather than more layoffs, and older workers in the same jobs show no similar decline.15

Census Bureau data point the same way. In the most AI-exposed fifth of industries, hiring of 22- to 24-year-olds fell about 9% soon after ChatGPT's release and has not recovered.9 A related study tracking 6.7 million bachelor's graduates found that those from the most exposed majors were 5 percentage points less likely to be employed right after graduation, with starting pay about 13% lower, which is comparable to graduating into a recession.9 The Dallas Fed concluded in September that generative AI has reduced employment and earnings for new graduates whose skills are easy to automate.10 A New York Fed survey of AI-using service firms adds detail: 15% said they had hired fewer people and only 4% reported AI-related layoffs.11

This is where Dixon's framing and the hiring research part ways, though they don't contradict each other. If AI mainly shows up as jobs that never get posted, most current workers won't experience it as a loss. A survey of people with jobs is close to the worst tool for seeing that effect. The "surprise" in his results is real for incumbents, but it should not be read as a clean bill of health for the labor market as a whole.

The international evidence is also split. Nordic studies using full-population records in Finland and Norway found little or no AI-linked drop in youth employment, and some found declines that began before ChatGPT.9 One U.S. study found that unemployment risk in AI-exposed jobs began rising in early 2022, before ChatGPT's release, which raises the possibility that interest rates or remote work explain some of what gets blamed on AI.9

The layoff narrative versus the layoff data

There is also a gap between what companies say and what workers report. Challenger, Gray & Christmas counted 120,136 announced U.S. job cuts attributed to AI through September 2026. That is about 21% of all announced cuts, making AI the most-cited reason this year, even as total announced cuts were down 39% from 2025.15 Dixon notes that when his survey went into the field, unemployment was 4.1%, the workforce was shrinking and AI was reportedly a leading reason for many layoffs.1

These figures are not the same as measured job losses. They reflect reasons employers chose to give for planned cuts, not independently confirmed layoffs.1115 Since companies may prefer to describe cost-cutting as modernization, the low share of workers reporting AI-driven job loss is a useful check on the corporate story. One labor-market analysis argues that AI's biggest cost so far may be slower wage growth for exposed workers rather than layoffs, putting the total at roughly $28 billion a year.12

Perception is running ahead of experience

The gap between what people fear and what they have experienced is wide. A Pew Research Center survey found that in 34 of 37 countries, people are more likely to expect AI to reduce jobs than to add them. In the U.S., about seven in ten adults expect job losses over the next 20 years.18 Official projections are more measured. The Bureau of Labor Statistics expects U.S. employment to grow by 5.9 million jobs over its 2025–35 projection period, while office and administrative support jobs shrink by 4.0% and data scientist jobs grow 34.6%.15

The bottom line

The best reading of the evidence is that Dixon is right about the people he measured, and those people are not where the strain is. For most Americans who have jobs, AI has so far meant new tools and, for some, a promotion, not displacement. The real pressure is at the entry point to the labor market, where fewer openings for young graduates are hard to see in surveys of the employed and easy to misread in corporate layoff announcements. Dixon says he is now looking at who is getting promoted and how workers expect AI to shape their future opportunities.3 Those answers, together with ongoing hiring data for young workers, will show whether today's limited impact turns out to be lasting or just a delay.