Texas graduates in more AI-exposed majors had lower first-year employment and pay. The study still cannot prove AI caused it.
A new Federal Reserve Bank of Dallas analysis joins Texas public-university education records to Texas employment and earnings records. It asks whether the first labor-market year changed differently for graduates from majors that historically feed into more generative-AI-automatable work.
After ChatGPT’s late-2022 release, a 10-percentage-point higher share of automatable tasks in a major was associated with a 1.7-percentage-point lower probability of Texas employment within a year of graduation. Among graduates who found work in Texas, first-year earnings were about 5% lower relative to the change for graduates from less-exposed majors. The authors report stable differences before ChatGPT.
What this adds
This is another observed early-career outcome, not a model or a task map alone. It comes from a different administrative dataset than the earlier multi-state Census graduate study, which also found weaker initial employment and earnings in more-exposed majors. The warning is therefore harder to dismiss as one data source.
What it does not establish
The exposure measure is built from pre-ChatGPT job-posting demand for each major and task scores based on Claude use. It does not show which employer deployed an AI system, when it did so or whether that use changed a specific job opening. The after-2022 difference could still partly reflect another shock that hit more-exposed majors. The records also cover Texas public-university graduates and Texas employment, not every graduate or job.
Why the unemployment check still matters
An IZA analysis of U.S. CPS data did not find a statistically significant summer-2026 unemployment spike among recent college graduates. That does not erase a first-job employment or earnings gap: unemployment, first-year job match and pay are different outcomes, measured in different cohorts and windows. It does mean this result is not a license for a national graduate-unemployment headline.
The honest current warning
The risk that is now measured most consistently is a weaker entry route in some exposed U.S. fields, not an established AI layoff wave. Watch for studies that follow the same graduates and firms while observing actual AI rollout, remote-work change, hiring, earnings and later job quality.