Is AI driving unemployment?

Since the launch of ChatGPT in November 2022 unemployment rates have risen across much of the industrial world. Some see this as the early stages of a period of rapid, technology-driven job destruction – one that will hit the labour market like a “tsunami”, according to Kristalina Georgieva, the IMF’s managing director.

It would be surprising if AI did not destroy some jobs. Most transformational technologies do. There’s never a time when some technology or innovation wasn’t reshaping work and destroying jobs. But technology also creates new types of work, and, by raising productivity and demand, fuels job growth elsewhere in the economy. Decades of rapid technological change have been accompanied by growing, not falling, employment. The number of US jobs has risen sixfold since 1900 while, in recent years, unemployment has run at exceptionally low levels.

Many worry that AI will break this pattern, leading to, in the words of London mayor Sadiq Khan, a “new era of mass unemployment”.  In this world a super capable intelligence could replace humans even in jobs that require creativity, empathy, judgement and adaptability – the sort of work that has benefited from previous waves of technological change.  

What do the data show? The launch of ChatGPT has, indeed, been followed by rising levels of unemployment in the US, Germany, France and many other industrialised countries. Moreover, a number of US and UK studies have shown a marked weakening in employment for younger workers in AI-exposed occupations, such as software engineers, since 2022.  A Stanford analysis of private sector payroll data found, “early, large-scale evidence consistent with generative AI disproportionately impacting entry-level workers in the American labor market”. (Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, Brynjolfsson et al, November 2025.) 

Is this really about AI? Much of the weakness in the job market can be explained by old-fashioned cost control, as companies have sought to counter the effects of the surge in inflation in 2021–22 and the ensuing rise in central bank interest rates in 2022–23. (The US price level today is 24% higher than in early 2021. US interest rates rose from 0.25% in early 2021 to 5.5% in the summer of 2022.) 

An analysis of millions of job ads by the FT’s John Burn-Murdoch found that the hiring slowdown started six months before the introduction of ChatGPT, in mid-2022, when US interest rates were rising, and US equities were falling. (The NASDAQ index of tech stocks fell by one-third between late 2021 and September 2022.) Higher costs and a major tightening of monetary policy hit corporates hard. Burn-Murdoch points out that the most AI-exposed sectors, such as tech and professional services, are also highly exposed to shifts in interest rates and external shocks. And in the case of tech, job losses seem at least in part an unwinding of the surge in hiring during and immediately after the pandemic.  

Nor are rising levels of graduate and youth unemployment the smoking gun that some believe. Unfortunately, young people are always the first to feel any weakening in the job market. When things start to cool, employers cut hiring first, hitting younger workers, while existing, older employees sit tight. Rates of youth unemployment are far more volatile than rates of unemployment for older workers.  

So even the squeeze on tech jobs and younger workers, which at first sight looks like clear-cut evidence of the impact of AI, can be explained by more prosaic factors – the surge in business costs, a sharp rise in interest rates and the equity sell-off. 

This is not to say that AI is having no effect on the job market, just that the effects are less pronounced than the job losses caused by older technologies that happen continuously. 

My colleague Tom Avis has looked at the change in the number of jobs in every UK sector over various two-year periods. Technology-led destruction can be seen everywhere, even as overall levels of employment have risen. Self-service checkouts mean that there are 9.4% fewer sales and retail assistants in British shops than in 2023. Online gambling cut the number of betting shop managers by 62% between 2022 and 2024. The migration to online banking reduced the number of bank and post office clerks by 28% between 2015 and 2017. 

Technology is one factor that drives the endless process of job creation and destruction, but so, too, is the economic cycle, innovation, competition, regulation and much more. The current downturn in the labour market has much more to do with the familiar, cyclical suspects, such as interest rates and inflation, than AI. 

AI could reshape the world of work. So far it is doing so in an undramatic fashion. A recent study by the Yale Budget Lab found that the rate of change in the mix of US jobs since the launch of ChatGPT has not been markedly different than that seen after the advent of the personal computer in the 1980s or the internet in the 1990s. Not so much a case of ‘this time it’s different’ as ‘this time it’s the same’.