One in five US job cuts now cites AI. The labour data has not caught up

I.The tally says AI. The trend says the wave has passed

On 1 October, Challenger, Gray & Christmas reported that US employers announced 43,281 job cuts in September, down 18% from August and 20% from a year earlier. AI was cited for 3,961 of them, fifth among stated reasons, behind market and economic conditions (8,789), closings, falling demand and restructuring. For the year to date, though, AI has been cited for 120,136 cuts, 21% of the total and more than any other reason.

Coverage of the release offered two different hiring figures, and both are right. Challenger's primary release puts September hiring plans at 90,787, 23% below the 117,313 of September 2025, while hiring plans for the year to date, at 210,612, are 3% above 2025. One figure covers a month and the other covers nine.

The 21% headline is a running total. Month by month, the picture is different. AI was the leading monthly reason for five months starting in March, but in August it fell to fourth place with 3,462 cuts, its lowest monthly total since December 2025. September was barely higher. The first half of the year accounted for 101,743 AI-cited cuts, about 85% of the year-to-date total.

II.AI went from footnote to leading reason in a year

The jump is real. In the first nine months of 2025, Challenger counted 17,375 cuts explicitly tied to AI, plus 20,219 under the broader heading of technological updates. The full-year 2025 total was 54,836, and 71,825 since 2023. The first nine months of 2026 therefore show about seven times the same period of 2025, and more than twice all of last year.

The cuts are concentrated in one sector. Technology companies announced 165,925 cuts this year, 29% of the total and 54% more than in the same period of 2025. Fast Company counts roughly 3,300 roles at Uber among September's announcements.

The 21% share is also partly an effect of a smaller total. Year-to-date cuts are down 39%, to 573,195, mainly because 2025 included the federal workforce reductions, which Challenger attributed to DOGE for 293,753 cuts over that year. Excluding government, cuts are down just 15%. Our own estimate: had 2026's total matched 2025's 946,426, the same 120,136 AI-cited cuts would be about 13% of it. The number of AI-cited cuts rose a lot, and the fall in total cuts makes the share look larger still.

III.An announcement is not a displacement

Challenger counts announced plans and records the reason the employer gives. A plan to cut jobs is not the same as people losing them, and the firm's December 2025 report sets out no method for checking whether the stated reasons are true. Andy Challenger has been open about the incentive. In August he said that naming AI "can win over investors", and that company messaging had moved from hedging to openly citing it.

The case that firms are exaggerating AI's role has strong supporters. In January, Oxford Economics found AI accounted for about 4.5% of reported US job cuts in the first eleven months of 2025. It suggested companies were presenting ordinary cuts, including corrections for past over-hiring, as innovation.

The more important point is scale. The BLS Job Openings and Labor Turnover Survey (JOLTS) recorded 1.6 million layoffs and discharges in August, a rate of 1.0%, alongside 5.2 million hires. Our estimate, assuming layoffs ran near that monthly level from January to August: about 13 million layoffs, of which the 120,136 AI-cited announcements would be under 1%. Most US layoffs are never announced as cuts, and the announced ones that cite AI are a small slice of those.

Studies that compare exposed and unexposed occupations find no effect so far. The Budget Lab at Yale, writing in May, found no statistically significant effect of AI on employment or real wages in AI-exposed occupations as of early 2026. The New York Fed's September survey of New York and northern New Jersey firms found 61% of service firms using AI, but only 4% had laid off workers because of it. A further 15% had reduced hiring, and 33% had retrained staff.

Two cautions apply. A finding of no significant effect does not show the effect is zero, and the New York Fed survey covers one region. Yale's own authors expect AI to leave its mark eventually. This site's August brief on the young-worker gap also found effects showing up first among entry-level workers. Both point the same way: the damage, where it exists, comes through jobs that are never posted rather than people who are let go.

IV.Where AI shows up: hiring, productivity and pay

Labour. The September jobs report showed payrolls up just 29,000, with unemployment at 4.2%, and July was revised to a loss of 10,000. Layoffs are not driving this. The layoff rate is 1.0%, and one analysis of JOLTS puts hires at 3.16 times layoffs, above the median since 2007. The weakness is in hiring. Employers are hiring and firing little, which fits the New York Fed finding that nearly four times as many service firms cut hiring as cut staff.

Productivity. Nonfarm business productivity rose 2.2% over the year to Q2 2026, with output up 2.5% and hours up only 0.2%. That is what a labour-saving technology would produce. It is also what a cyclical hiring freeze produces. The productivity slowdown Oxford Economics pointed to in January has since faded, so that argument against AI-driven change is weaker now than it was.

Capital versus labour. In the same release, the labour share of output fell to 52.8%, the lowest since the series began in 1947, and real hourly compensation fell 0.1% over four quarters. Output is rising while workers' share of it shrinks. That gap is where AI's economic effect would show up first, before any wave of layoffs. The BLS data cannot separate AI from other causes, and we do not claim it does.

Market structure. AI-cited cuts are concentrated in large, investor-facing tech firms, which have the most to gain from presenting cost cuts as innovation. Rewarding that framing pushes firms to restructure around automation narratives before the technology is ready. That favours incumbents who can afford to restructure early.

Prices. With real pay roughly flat and September hiring plans 23% below a year earlier, the more likely near-term effect is weaker household demand, not faster wage inflation. This is our analytical estimate, not a measured effect.

V.Second-order: policy built on the label measures the narrative

If retraining money, disclosure rules or taxes are keyed to the reason a company gives for cuts, they will track how firms describe their cuts, not what AI is doing to jobs. Andy Challenger has already warned that as regulation takes shape, companies will be more careful in their announcements and the AI signal will become harder to read. A falling AI tally in 2027 could mean firms have stopped saying it, not that displacement has stopped.

Workers are also reacting to the headline. The New York Fed's August analysis notes that workers exposed to generative AI hold bleaker expectations for jobs and income. It also reports that firms expect to cut hiring further, particularly for college-educated workers. A widely repeated layoff figure can change how people choose careers and how hard they bargain before it changes employment.

VI.Precedent: offshoring also looked small in the layoff counts

In the first quarter of 2004, when offshoring dominated US politics, BLS mass-layoff data attributed 4,633 of 239,361 separations to overseas relocation, about 1.9%. Years later, Autor, Dorn and Hanson showed that import competition from China caused large, lasting job losses in exposed local labour markets.

The lesson cuts both ways. Counts of stated reasons can exaggerate a shock, as AI's 21% may, or understate one, as offshoring's 2% did. In both cases the real effect showed up slowly, in particular places and among people who were never hired, more than in the headline layoff count.

VII.How it plays out: three scenarios

Probabilities are our analytical judgment, not model forecasts, and they sum to 100%.