Contents
Few questions about technology cause more worry than this one. Headlines warn of the end of the graduate job, and surveys find many young people believe their prospects have already shrunk. Yet nearly four years after ChatGPT was released, the best data on jobs, pay and hiring tell a more complicated story than either the fears or the reassurances. This deep dive sets out what has been measured, what has only been forecast, and where the first real signs of change are appearing.
Chapter 1
Why this is hard to measure
Working out what AI is doing to jobs is genuinely difficult, for three reasons.
Many things are happening at once. Since 2022 the economies of the US and the UK have been through high inflation, rising and then falling interest rates, and a slowdown in hiring. Any of these can cut job openings. Separating the effect of AI from the rest needs careful methods.
Exposure is not the same as replacement. Researchers usually start by asking which occupations involve tasks that AI can do, which they call exposure. But a job can be exposed because AI takes over part of it, or because AI makes the people doing it more productive. Researchers at the Budget Lab at Yale point out that if half of exposed jobs grew and half shrank, the average effect could look like zero.
Change shows up first in hiring, not firing. Companies rarely sack staff the moment a new tool arrives. More often they hire fewer people to replace those who leave, or fewer newcomers. That kind of change is slow to appear in employment totals.
Chapter 2
What the forecasts say
Before looking at what has happened, it helps to see the scale of what is expected.
| Source | What it measures | Headline finding |
|---|---|---|
| IMF, 2024 | Share of jobs exposed to AI | Almost 40% worldwide, about 60% in advanced economies |
| ILO, 2025 | Share of workers in jobs with some generative AI exposure | One in four worldwide, 34% in high-income countries |
| World Economic Forum, 2025 | Jobs created and lost from all major trends by 2030 | 170 million created, 92 million displaced, a net gain of 78 million |
The International Monetary Fund’s analysis, presented by its managing director Kristalina Georgieva, found that almost 40% of jobs worldwide are exposed to AI, rising to about 60% in advanced economies such as the UK. It expected roughly half of those exposed jobs to benefit, through higher productivity, and the rest to face lower pay and fewer openings. It also warned that in most scenarios AI is likely to worsen inequality overall.
The International Labour Organization’s 2025 index uses a more detailed, task-by-task method. It finds that one in four workers globally is in an occupation with some exposure to generative AI, but only 3.3% are in the most exposed category. That top category is more common among women, at 4.7% of female employment against 2.4% of male employment. The ILO’s conclusion is that transformation of jobs is the most likely impact.
The World Economic Forum’s Future of Jobs Report 2025, based on a survey of more than 1,000 employers, projects 170 million jobs created and 92 million displaced by 2030, a net gain of 78 million. Those figures cover all the big forces at work, including the green transition and demographic change, not AI alone. The same report expects nearly 40% of the skills needed in jobs to change.
All three are projections. They describe where AI could reach, not what it has done.
Exposure is not the same as replacement. A job can be exposed because AI takes over part of it, or because AI makes the people doing it more productive.”
| Item | % of employment |
|---|---|
| Advanced economies | About 60% |
| Global | Almost 40% |
| Emerging markets | 40% |
| Low-income countries | 26% |
Exposure means AI could affect the job, for better or worse.Source: International Monetary Fund. Chart: The Sidelight.
Chapter 3
What has happened so far: little sign of overall change
The most careful studies of the economy as a whole have found little so far.
In the US, the Budget Lab at Yale has been tracking the labour market since ChatGPT’s release. Its May 2026 analysis compared the employment and wages of AI-exposed occupations with similar unexposed ones, using a method designed to strip out other economic changes. It found no statistically or economically significant effects so far. The authors added that this could change quickly.
Denmark offers an unusually detailed view, because surveys of which workers use AI chatbots can be linked to official records of their pay and hours. A study by the economists Anders Humlum and Emilie Vestergaard, published by the US National Bureau of Economic Research, found no effect of AI chatbots on earnings or recorded hours. It was precise enough to rule out changes larger than 2%, two years after ChatGPT’s launch.
These results fit the evidence from workplace experiments, discussed in our article on whether AI makes workers more productive. Most found that AI speeds up some tasks, with the largest gains for less experienced workers. Faster tasks have not yet translated into visible changes in jobs or pay across whole economies.
Chapter 4
The warning sign: young workers
There is one place where a change does show up, and it is among people starting their careers.
Economists at Stanford’s Digital Economy Lab studied payroll records covering millions of US workers, supplied by the payroll company ADP. Their paper, titled “Canaries in the Coal Mine?” and most recently updated in August 2026, makes several findings.
- No economy-wide displacement. Across the workforce as a whole, they found no evidence of widespread job losses.
- A gap for young workers. In the occupations most exposed to AI, employment of 22- to 25-year-olds is now 19% below where it would have been had it kept pace with their peers in less exposed jobs. The gap has widened steadily since August 2025.
- No gap for experienced workers. Older workers in the same occupations show no comparable fall.
- Fewer hires, not more firings. The difference comes mainly from companies hiring fewer young people, not from sacking those already employed.
- It depends how AI is used. The falls are concentrated in jobs where AI mainly substitutes for human tasks. Where AI mainly helps workers do their jobs, employment is flat or rising.
The authors are careful about what this proves. They describe their findings as early warning indicators, not proof of cause. The pattern holds when they exclude technology companies and adjust for interest rates and remote working, but it weakens when they account for education, and some of the trends began before generative AI arrived. The gap is also larger in their payroll data than in national surveys.
Still, the pattern is consistent with how many expected AI to arrive: not as mass redundancies, but as a quieter decision to hire fewer people at the bottom of the ladder.
The pattern is consistent with how many expected AI to arrive: not as mass redundancies, but as fewer people hired at the bottom of the ladder.”
Chapter 5
The UK picture
The UK has no study as detailed as the American ones, and the evidence is mixed.
Graduate hiring has weakened. But Charlie Ball, a labour market specialist at the graduate careers service Prospects, argued in an analysis for the Institute of Student Employers in June 2026 that the official figures do not show a collapse. Graduate outcomes were a little worse than the year before, and the share of employed graduates in professional-level jobs dipped from 71.9% to 70.9%.
Some occupations that could be affected by AI, such as coding and management consultancy, did take fewer graduates. But some of the biggest falls were in teaching, housing and human resources, and in public sector employers. Ball suggests that points to cuts in public sector hiring more than to AI. Meanwhile medicine, nursing, accountancy and law took on more graduates. He found no compelling evidence that AI is commonly replacing graduate jobs, and described the picture as evolution, not elimination.
Young people themselves are more worried. In a survey for Prospects, 35% of students and graduates said AI had already reduced their career opportunities.
The wider youth jobs market is weak whatever the cause. The Office for National Statistics estimates that 981,000 people aged 16 to 24 were not in education, employment or training in April to June 2026, or 13.0% of the age group, up 30,000 on a year earlier. Those figures do not separate out the effect of AI.
Chapter 6
Reading the evidence together
| Study | Where | Finding |
|---|---|---|
| Yale Budget Lab, May 2026 | US | No clear effect yet on employment or wages in AI-exposed jobs |
| Humlum and Vestergaard, 2025 to 2026 | Denmark | No effect on earnings or hours, ruling out changes above 2% |
| Stanford Digital Economy Lab, 2026 | US | Employment of 22- to 25-year-olds in AI-exposed jobs 19% below trend; no gap for older workers |
| Ball for ISE, June 2026 | UK | Graduate market weaker but not collapsing; falls not concentrated in AI-exposed fields |
These findings are not as contradictory as they first look. An economy-wide average can stay flat while one group, such as young people entering exposed occupations, is hit. A hiring slowdown for new entrants takes years to show up in total employment, because most people in a job keep it. And the UK and Danish studies use different data, over different periods, from the US ones.
The most defensible summary is this: there is no evidence yet of large-scale job losses from AI, and some evidence that it is starting to reduce opportunities for young people in the most exposed jobs, at least in the US.
Chapter 7
What would change the picture
Several developments would show that AI’s effect on jobs is growing.
- The young-worker gap spreading. If the fall in hiring of young people extends to older workers, or to less exposed occupations, the effect is broadening.
- Yale’s method detecting an effect. The Budget Lab’s approach is built to pick up a signal once one appears. A finding of significant effects there would carry weight.
- Effects on pay. So far the Danish data show none. Falling pay in exposed jobs would be an important change.
- The more capable AI agents now arriving. Much of the evidence comes from a period when AI mainly meant chatbots that assist people. Systems that complete whole tasks on their own could have a larger effect, and the data may take time to show it.
The opposite would also be informative. If exposed occupations grow, as the IMF expects for about half of them, the early signs of lost entry-level hiring could prove to be a pause while companies work out how to train new staff alongside AI.
Chapter 8
What it means for workers
The evidence does not support panic, and it does not support complacency.
- Early-career workers face the most uncertainty. The only clear hiring gap in the data so far is for young people entering AI-exposed jobs.
- Using AI well seems to matter. Studies find employment holding up or rising where AI helps workers, and falling where it does the work instead.
- Forecasts point to change more than loss. The IMF, ILO and World Economic Forum all expect most affected jobs to be reshaped. The WEF expects nearly 40% of job skills to change by 2030, which puts a premium on training.
The honest answer to the question is that AI is not yet taking jobs on a large scale, and that the first group it does seem to be affecting is the one with least experience to fall back on.
References and sources
- 01Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence · Brynjolfsson, Chandar and Chen, Stanford Digital Economy Lab · Revised 12 August 2026
- 02What We Do and Don’t Know About How AI is Affecting the Labor Market · Gimbel, Kendall and Nunn, The Budget Lab at Yale · 7 May 2026
- 03Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI · Humlum and Vestergaard, National Bureau of Economic Research · May 2025, revised March 2026
- 04AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity. · Kristalina Georgieva, International Monetary Fund · 14 January 2024
- 05Generative AI and Jobs: A Refined Global Index of Occupational Exposure · International Labour Organization · 20 May 2025
- 06Future of Jobs Report 2025: 78 Million New Job Opportunities by 2030 · World Economic Forum · 7 January 2025
- 07The graduate jobs market isn’t collapsing · Charlie Ball, Institute of Student Employers · 17 June 2026
- 08Young people not in education, employment or training (NEET), UK: August 2026 · Office for National Statistics · 27 August 2026
- 09Lead image: Graduation (UHPC 6212).jpg · Unknown photographer, via Wikimedia Commons (public domain) · Accessed 13 November 2026
House byline
Articles under this byline are researched and drafted with AI assistance, with every figure checked against its source before publication. The editor, Pankaj Guleria, reviews them and corrects them where needed.
Edited by Pankaj Guleria, Founder and Editor



