The answer in four numbers
Two for the whole economy, two for the workers most exposed to AI. Each opens to the monthly data behind it and its source.
Whole economy
Unemployment rate
Oct 2026, all US workers, seasonally adjusted.
Whole economy
AI-cited share of layoffs
Jan to Sep 2026: 120,136 AI-cited cuts against 15.6 million layoffs and discharges.
Most exposed
Recent college graduates
Unemployed in Sep 2026, ages 22 to 27 with a degree, against 4.0% for all workers.
Most exposed
Information industry
Unemployed in Oct 2026, in the industry using AI most. Not seasonally adjusted.
Recent graduates are now more likely than the average worker to be unemployed
For most of the past 36 years, a new degree meant a lower chance of unemployment than the average worker had: in January 1990, recent graduates' rate was 1.8 points below the average. That flipped in January 2021, before ChatGPT's release in November 2022, so AI cannot be the whole story. The gap has widened since, from 0.6 points when ChatGPT came out to 1.3 points in September 2026.
Unemployment by type of work
| Occupation group | A year earlier | Oct 2026 | Change |
|---|---|---|---|
| Management, professional and relatedThe white-collar work most exposed to AI | Not published | n/a | |
| Office and administrative supportClerical work, among the most automatable | Not published |
Why nobody can count AI unemployment directly
- The unemployment survey does not ask why.
- Unemployment is measured by a monthly household survey that asks whether people worked and whether they looked for work. It does not ask what cost them their last job, so there is no official count of people out of work because of AI, or because of any other single cause.
- A job cut is not an unemployed person.
- Challenger and this site's register count the job cuts companies announce. Many cuts are spread over months or absorbed by people leaving on their own, and many people who lose a job find another before they show up in unemployment for long.
- Most layoffs are never announced.
- US employers carried out about 1.7 million layoffs and discharges a month from January to September 2026, and Challenger's count of announced cuts for every reason equals about 4% of them. Any count built from announcements sees only a slice of all job losses, so AI's share of announcements says little about AI's share of job loss.
What would show AI is starting to raise unemployment
Four signals, each with its latest reading. If AI begins to push unemployment up, it should show here first.
- AI-cited cuts becoming a meaningful share of all layoffs0.8%Jan to Sep 2026, against 0.3% in 2025
- The recent-graduate gap widening further+1.3 ptsSep 2026, against +1.7 pts a year earlier
- AI-attributed cutting staying high against its own history, as measured by the AI Layoffs Index58.4 of 100Elevated, recomputed Nov 18, 2026
For the count of AI layoffs themselves, see how many jobs AI has replaced, and for the companies behind them, every company in the register.
Frequently asked
Is AI causing unemployment?
Not in the economy-wide numbers, at least not yet. The US unemployment rate was 4.1% in October 2026. Job cuts that US employers attributed to AI from January to September 2026 equal less than 1% of all layoffs and discharges over the same months. The strain that does show is narrower: recent college graduates were unemployed at 5.2% in September 2026, against 4.0% for all workers, a gap that is narrower than a year earlier.
See the four numbersHow many people are unemployed because of AI?
No official statistic says. The government's unemployment survey asks whether people are working and looking for work, not which technology cost them their last job, so nobody can count AI unemployment directly. The closest measures count job cuts, not unemployed people: Challenger, Gray & Christmas counted 121,401 US job cuts that employers attributed to AI in 2026 through October, and across the layoffs the press linked to AI, employers themselves named AI for 72,868 roles worldwide. Neither is a count of people still out of work, because many people who lose a job find another.
How many jobs has AI replaced?