NBER: AI writing in dissertations predicts industry careers

Nearly three in ten US STEM doctoral abstracts filed in early 2026 contain AI-generated text, according to a new working paper that scanned more than 226,000 dissertations completed between January 2019 and May 2026. In 2023 the share was 1.7%. In 2025 it was 18.8%. By the first five months of 2026 it had climbed to 29.4%, Nature reports.

The more striking finding is downstream. Among US-educated PhD recipients whose dissertation abstracts tripped an AI-writing flag, graduates were 17% less likely to work in academia and 16% more likely to work in industry than classmates in the same program and cohort. "Among graduates in the same program and cohort, those whose dissertations contain AI writing are less likely to enter academic research careers and more likely to enter non-tenure-track roles or industry," the paper reports.

Use is not spread evenly. Mathematical and computer sciences dissertations led at 36%, engineering at 35%, physical sciences trailed at 22%. Graduates educated in non-English-speaking countries were flagged at 41%; those from English-speaking countries at 23%. The authors (Daniel P. Gross, Dror Shvadron and Hansen Zhang) operationalized AI use as containing at least 2% AI-written text, identified using the Pangram 4.0 detection model.

Gross framed the concern bluntly: "If our students are only doing half the work and they're dishing the rest to AI, do they really have the depth of understanding" to be effective scientists?

NBER Working Paper 35859 remains unreviewed. Two researchers we follow passed the piece along within days of its publication.