No major measurement supports the claim that AI will replace most jobs. The International Labor Organization (ILO) estimated in 2025 that about one in four workers worldwide is in an occupation with some exposure to generative AI, and only 3.3 percent of global employment sits in its highest exposure gradient. The ILO concludes that transformation of jobs is the likelier outcome than wholesale replacement, because most occupations still contain tasks that need human input.
Key takeaways
- The ILO's 2025 index puts 34 percent of employment in high-income countries and 11 percent in low-income countries in occupations with some generative AI exposure.
- The International Monetary Fund (IMF) said in January 2024 that almost 40 percent of global employment is exposed to AI and that roughly half of the exposed jobs in advanced economies may see key tasks done by AI.
- Economists who traced new job titles from 1940 to 2018 found that most current employment is in specialties introduced after 1940, and that automation slows the creation of new work.
- Yale's Budget Lab tracker, updated September 15, 2026, found no clear evidence of labor market disruption tied to AI.
- Exposure counts tasks that software could touch. It does not count jobs lost.
What "most jobs" would have to mean
The phrase hides three different claims. One is about exposure, meaning how much of a job's task list a system could assist or perform. A second is about use, meaning what workers and firms actually do with the tools. A third is about displacement, meaning fewer people employed. Most alarming headlines quote the first and imply the third. The table lines up what the sources reviewed here actually measure against what a "most jobs replaced" claim would need.
| Source | What it measures | Headline figure | Does it show replacement? |
|---|---|---|---|
| ILO working paper 140, May 20, 2025 | Task exposure to generative AI by occupation, worldwide | About one in four jobs with some exposure; 3.3 percent in the highest gradient | No. The authors expect transformation more than replacement |
| IMF blog, January 14, 2024 | Share of jobs exposed, by income group | Almost 40 percent worldwide; about 60 percent in advanced economies | No. About half of exposed jobs may benefit, the rest may lose key tasks |
| NBER 30389, published in the Quarterly Journal of Economics in 2024 | Origins of new job titles, 1940 to 2018 | Most current employment is in post-1940 specialties | Not as a forecast. It describes how new work offset past automation |
| Yale Budget Lab tracker, updated September 15, 2026 | Occupational churn and AI usage in US monthly survey data | No clear sign of AI-related disruption | No evidence found so far |
The ILO index in detail
The ILO's refined global index rests on a survey of about 1,640 workers, roughly 52,500 data points on the automation potential of about 2,900 tasks, and an AI model that scored the tasks in the international occupation classification. Clerical occupations remain the most exposed, and some highly digitized professional and technical roles have become more exposed than in the 2023 version. Exposure is higher in richer economies: 34 percent of employment in high-income countries against 11 percent in low-income countries. Within the highest gradient, 4.7 percent of female employment is affected against 2.4 percent of male employment.
Reading the IMF's larger number
The IMF's figure of almost 40 percent of global employment is larger than the ILO's one in four, and the two should not be added or averaged, because each agency built its own method and published its own scoring. Both are exposure measures. The IMF splits advanced-economy exposure into two halves: roughly half of exposed jobs may benefit from AI integration and higher productivity, and for the other half AI applications may carry out key tasks now done by people. That is a statement about tasks. Neither agency's text reviewed here states a count of jobs lost or a date by which a share of jobs disappears.
What earlier automation did to the job list
A long-run economic study gives the historical comparison. David Autor, Caroline Chin, Anna Salomons, and Bryan Seegmiller built a database of new job titles from 1940 to 2018 and linked it to Census microdata and patent-based measures. The working paper, later published in The Quarterly Journal of Economics, finds that most current employment is in specialties introduced after 1940. It also finds that technologies that complement an occupation tend to raise labor demand, while automation lowers it, and that the demand-reducing effect of automation has grown over the last four decades. The study is a record of the past, so it supports caution about both extremes: new work has appeared before, and automation has had measurable costs for specific occupations.
Checking the latest evidence
The Yale Budget Lab published its tracker on July 16, 2026 and updated it on September 15, 2026 using August 2026 US survey data. It reported that churn across occupations, AI exposure among unemployed workers, and AI usage data stayed flat, within historical ranges, or on pre-AI trends. The lab calls the result provisional and says it will keep monitoring. A short observation window cannot rule out later effects, and data from the United States do not describe Canada.
Questions to ask of any "most jobs" claim
- Is the number an exposure share, a usage rate, or a count of lost jobs?
- Does the study say which year, which AI tools, and which country it covers?
- Does it consider new occupations as well as lost ones?
- Is a date attached to the prediction, and who made it?
This article summarizes published research and is general information, not career or financial advice.
FAQ
Will AI replace most jobs?
None of the sources reviewed here forecasts that. The ILO expects transformation more than replacement, and the IMF describes exposure, with about half of exposed jobs in advanced economies possibly benefiting.
How many jobs are exposed to generative AI?
The ILO's 2025 estimate is about one in four jobs globally. The IMF's 2024 estimate is almost 40 percent of global employment. The two differ because they define and score exposure differently.
Are women or men more exposed?
In the ILO index, 4.7 percent of female employment and 2.4 percent of male employment worldwide fall in the highest exposure gradient. The ILO does not tie this to any forecast of job losses.
Has AI already caused mass unemployment?
The Yale Budget Lab tracker found no clear evidence of AI-related labor market disruption in US data through August 2026. The result is provisional.
Related reading
- US and Canadian labor data on AI jobs
- AI and the work of scientists
- AI and white-collar work
- Whether AI could replace teachers
More in AI and Society.
Last verified: October 2026
Sources
- Generative AI and Jobs: A Refined Global Index of Occupational Exposure, ILO Working Paper 140, ILO, United Nations labor agency, May 20, 2025
- AI Will Transform the Global Economy. Let's Make Sure It Benefits Humanity, IMF Blog, International Monetary Fund, January 14, 2024
- Autor D, Chin C, Salomons A, Seegmiller B, New Frontiers: The Origins and Content of New Work, 1940-2018, NBER Working Paper 30389, National Bureau of Economic Research, 2022, published in The Quarterly Journal of Economics, 2024
- Tracking the Impact of AI on the Labor Market, The Budget Lab at Yale, published July 16, 2026, updated September 15, 2026