Research update · 22 September 2026
The broad U.S. data still do not show an AI jobs collapse.
Yale Budget Lab’s update with August CPS microdata finds no clear AI-related labor-market disruption in the broad measures it tracks. That is a serious counter-signal to a national layoff headline. It is not proof that every entry route or occupation is safe.
What the tracker measures
The tracker follows occupational churn, AI exposure among unemployed workers, AI-use measures and a synthetic difference-in-differences comparison by occupational exposure. Yale says these measures are flat, within historical ranges or continue along pre-AI trends through August 2026.
The result concerns broad U.S. indicators. It does not measure a worker’s personal risk, a named employer’s adoption decision or every dimension of job quality.
The correct headline is “not clear yet,” not “nothing is happening”
No clear footprint in an aggregate monitor is evidence against a claim that AI has already caused a nationwide jobs collapse. It is not a clean causal estimate: the tracker does not randomly assign AI use, and observed labor change has many possible causes.
The update also cannot prove the absence of a smaller effect. An aggregate indicator can stay stable while a particular entry route, geography or field changes. The timing, scale and persistence of any such effect remain open.
Why the counter-signal does not erase the entry-level warnings
U.S. studies have reported weaker early-career hiring and lower initial employment and earnings in more AI-exposed graduate pathways. A Texas analysis has found a relative pullback in online postings for more automatable work. Those studies target different populations, locations and outcomes from Yale’s broad tracker, and most use exposure rather than observed employer adoption.
The evidence therefore does not add up to one verdict. Broad U.S. disruption is not clear. Concentrated warnings remain plausible but unsettled. A common longitudinal study must test broad and subgroup outcomes together with actual adoption before either conclusion can crowd out the other.
What changes in the cumulative answer
The pool now has a current national counter-signal that is stronger than a slogan. Yale’s tracker, updated through August, does not find a clear broad AI labor-market footprint. This narrows what the pool can say: no evidence here supports “AI has already wiped out U.S. jobs.” It does not authorize “AI cannot hurt my job.”