Research check · 16 October 2026
AI-written dissertations track a different career path, not a count of lost jobs
An NBER working-paper abstract reports that likely AI-written text appeared in 29% of U.S. STEM PhD dissertations filed in 2026. Graduates with detected AI writing were less likely to enter academic research careers than graduates in the same program and cohort. The comparison does not show that AI writing caused the difference.
What the study measured
Gross, Shvadron and Zhang studied STEM PhD dissertations filed in the United States from 2019 to 2026 and linked graduates to later employment. The abstract reports no detected AI writing before 2023 and detection in 29% of dissertations filed in 2026. Graduates with detected AI writing were more likely to take non-tenure-track or industry roles and less likely to enter academic research, especially among U.S.-origin graduates.
Where the claim stops
AI writing was detected in text; the public abstract does not establish each student's actual use or why they used it. Students already planning industry careers may write differently, and language background differs across groups. We could not inspect the full paper's detector checks, exact effect sizes or model details. Industry work is employment, not an AI-caused layoff. This narrow U.S. doctoral sample cannot describe the wider job market.
Does this change the answer?
No. The study raises a useful question about how AI-assisted work may interact with training and career choices. It does not resolve whether employer AI adoption is cutting hires, pay or headcount. A stronger test would measure assistance directly, prior career intentions and later skills alongside job outcomes.