NBER: PhDs using AI 17% less likely to stay in academia

Roughly 30% of US STEM PhD dissertations filed in the first five months of 2026 contain AI-generated text in their abstracts, and the graduates whose work was flagged are measurably less likely to end up in academia, Nature reported this week. The underlying NBER working paper, by Daniel P. Gross with co-authors Dror Shvadron and Hansen Zhang, puts the career gap at 17% less likely to be employed in academia and 16% more likely to work in industry.

The team ran the Pangram AI-detection service over more than 226,000 STEM dissertations completed between January 2019 and May 2026, and classified roughly 90% of flagged abstracts as entirely AI-generated. Prevalence climbed from 1.7% in 2023 to 18.8% in 2025 before hitting 29.4% in early 2026. Mathematical and computer sciences led at 36%, engineering at 35%, and physical sciences trailed at 22%.

The gap by linguistic background is wide: 41% of abstracts from graduates of non-English-speaking countries were flagged, against 23% from English-speaking countries.

'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 good scientists?' Gross, an NBER economist based in Durham, asks.

Ang Liu of the University of New South Wales told Nature that AI tools help where substantial knowledge already exists but become problematic in areas with incomplete or uncertain information. The paper reports a correlation; it does not establish that AI drafting itself drives graduates out of academic careers.