NBER: AI dissertation use linked to leaving academia

Roughly 29% of US STEM PhD dissertations completed in the first five months of 2026 contained AI-generated text, according to a new NBER working paper covered in Nature. In 2023, the figure was 1.7%.

The paper by Daniel P. Gross, Dror Shvadron and Hansen Zhang ran the AI-detection service Pangram across more than 226,000 dissertations submitted between January 2019 and May 2026. "90% of the flagged AI-generated abstracts were classified by Pangram as entirely AI-generated," said Gross, an NBER economist. Mathematical and computer sciences showed the highest usage at 36%. Among graduates from non-English-speaking countries, 41% of dissertations contained AI text, versus 23% from English-speaking countries.

Among US-origin PhDs, those with AI-generated abstracts were 17% less likely to work in academia and 16% more likely to work in industry than peers whose dissertations contained no AI text. The authors write that AI "thus appears to also be displacing a potentially important portion of the work" through which expertise develops.

Ang Liu, an engineer at the University of New South Wales, told Nature that "rigorous research training becomes even more important" when students work in areas where knowledge is incomplete.

The study is a working paper, not peer-reviewed, and the detection pipeline is a single proprietary tool. Two researchers we track posted the paper within hours of its release.