NBER paper: AI-written dissertations tilt PhDs away from academia

Roughly 29% of US STEM PhD dissertations filed in 2026 contained AI-generated writing, up from essentially zero before 2023, according to a new NBER working paper covered by Nature. The career correlation is the sharper finding: among graduates in the same program and cohort, US-origin PhDs with AI-written abstracts were 17% less likely to pursue academic employment and 16% more likely to work in industry.

The paper, by economists Daniel P. Gross, Dror Shvadron, and Hansen Zhang, examined more than 226,000 STEM dissertations completed at US universities between January 2019 and May 2026, flagging abstracts with the Pangram AI-detection service.

Use was not evenly distributed. 41% of dissertations by graduates from non-English-speaking countries contained AI-generated abstracts, against 23% from English-speaking countries, and the practice ran higher at lower-ranked programs.

"Large language models are diffusing rapidly into science, raising concerns that delegating research writing to AI may undermine the development of expertise that scientific training is supposed to build," the authors write. The tooling, they argue, "appears to also be displacing a potentially important portion of the work through which scientific expertise is developed."

The paper has not been peer-reviewed, and the career result is correlational, not causal: students more comfortable offloading writing may already have been less committed to the academic track. Two of the researchers we follow in our Who's Who directory shared the Nature piece within hours of its posting.