Artificial intelligence is making it harder for people to get a job, but it is not yet pushing large numbers of workers out of work. That is the main finding of a new study from the Federal Reserve Bank of Cleveland, published on 9 November 2026.
The researchers, Alexander Cline and Barış Kaymak, tracked American workers aged 23 to 54 month by month through the US Census Bureau’s Current Population Survey. They sorted 923 occupations into 20 groups, ranked by how much of each job’s tasks AI can now do, and compared the period since ChatGPT became widely used with a pre-AI baseline of 2015 to 2019.
The headline result is clear. Since generative AI became widely available, unemployment has risen by roughly one percentage point more in the occupations most exposed to AI than in those least exposed. The gap is even wider when the researchers use their most detailed measure of AI exposure: a predicted difference of 1.3 percentage points between the least and most exposed jobs.
Hiring slows, but layoffs do not rise
The study’s most important finding is about why unemployment has risen. It is not because firms are sacking more people. The researchers found no consistent evidence that workers in AI-exposed occupations are losing their jobs at a faster pace than before.
Instead, the problem is on the hiring side. People who are already out of work in AI-exposed occupations are taking longer to find a new job. Job-finding rates fell by roughly six percentage points more in the most exposed occupations than in the least exposed ones. Against a normal monthly job-finding rate of about 34 per cent, that shift means the average spell of unemployment stretches from around 10.5 weeks to more than 13 weeks.
The authors call this a “no hire, no fire” market. Firms are responding to AI by slowing new hiring and letting staff leave naturally, rather than by cutting jobs in large waves. A separate analysis from the Federal Reserve Bank of Dallas, published in August 2026, found the same pattern: lower employment for young workers in high-exposure jobs was driven mainly by fewer people being hired, not by more people being sacked.
Young workers feel it most
The pain is not spread evenly. Workers aged 23 to 30 face the sharpest gap. Moving across the range of AI exposure raises predicted unemployment for this age group by about two percentage points. For workers aged 45 and older, the gap is at most one percentage point.
The reason, the authors suggest, is that AI tools first replace routine thinking tasks, and those tasks sit heavily in entry-level roles. Young workers compete for exactly those starter jobs, so they feel the hiring slowdown first. The study describes this as a gradual restructuring of jobs “from the bottom up”.
That pattern is already visible in the wider jobs market. In the UK, vacancies have been falling for months, as reported in UK Jobs Market Cools as Vacancies Fall to 691,000. In the United States, employers added 165,000 jobs in October 2026 and unemployment edged down to 4.1 per cent, yet weekly jobless claims have stayed near historic lows — 200,000 in the week ending 7 November, the ninth straight week at or below that level, according to the US Department of Labor.
What the study does not prove
The researchers are careful about the limits of their work. They describe it as an “early and necessarily incomplete look” at a long change. The links they find are based on observation, not proof of cause and effect. And the survey data records the last job of an unemployed person, not the job they are now seeking, so first-time job seekers are undercounted.
There are also counterweights. The fastest job growth is happening outside software, and hiring in AI-heavy industries has rebounded. Experienced workers in exposed occupations have largely kept their jobs, and the Dallas Fed found their wages have been rising even as entry-level hiring shrank.
What this means for job seekers
For anyone looking for work, the message is practical rather than alarming. The risk from AI so far is concentrated at the moments of change: finishing education, switching jobs, or returning to work after a break. Workers already inside a job have, on this evidence, mostly kept it.
That makes early planning more valuable than panic. Job seekers may need to allow more time for a search, apply more widely, and build skills that AI cannot easily copy — such as judgement, face-to-face work and hands-on tasks. Employers, meanwhile, may need to rethink entry-level roles rather than simply freezing them, or the next generation of workers will find the first rung of the ladder harder to reach.
Sources: Federal Reserve Bank of Cleveland, “Job Finding and Unemployment in the Age of AI” (Economic Commentary 2026-26, 9 November 2026); Singularity.Kiwi, “The Fed’s New ‘No Hire, No Fire’ Study”; US Department of Labor weekly jobless claims report, 12 November 2026.
Comments (0)
Be respectful — comments are moderated.