NBER: AI-written PhD abstracts predict industry over academia

Roughly 29% of US STEM dissertation abstracts filed in the first five months of 2026 contained AI-generated text, up from 1.7% in 2023 and essentially nothing before that. That is the headline figure from a working paper by Daniel P. Gross, Dror Shvadron and Hansen Zhang, released by the National Bureau of Economic Research and covered by Nature.

The authors scanned more than 226,000 US STEM dissertations completed between January 2019 and May 2026, flagging any abstract where the Pangram AI-detection service rated at least 2% of the text as machine-written. Prevalence climbed from 1.7% in 2023 to 18.8% in 2025 before reaching 29.4% in early 2026. Mathematical and computer sciences led at 36%, engineering at 35%, and physical sciences trailed at 22%.

The share also split by linguistic background: 41% of dissertations by graduates from non-English-speaking countries carried an AI-flagged abstract, against 23% from English-speaking countries.

The career finding is what gives the paper its sharpest edge. Among US-origin graduates, those whose dissertations were flagged were 7.4 percentage points less likely to be employed in academia in 2026 (a 17% drop against a 44% base rate) and 8.0 percentage points more likely to be in industry (a 16% lift over a 50% base). The paper reports no difference in government employment.

"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 authors write.

The paper does not establish causation. The correlation could run through selection, with industry-bound students comfortable leaning on tools their academic-track peers still avoid. The authors disclose that they used Claude in their own initial drafting and that the AI text was "substantially revised."