NBER: AI-written PhD dissertations tied to academia exit

Among US-educated STEM PhD recipients whose dissertation abstracts contain AI-generated text, 17% fewer end up working in academia and 16% more go into industry, according to Nature's coverage of a new NBER working paper. The paper has not been peer-reviewed.

The team ran dissertation abstracts from more than 226,000 US STEM doctoral graduates completed between January 2019 and May 2026 through Pangram, an AI-detection service. AI-generated text first showed up in 2023 at 1.7% of abstracts, climbed to 18.8% in 2025, and hit 29.4% for dissertations filed in the first five months of 2026. Mathematical and computer sciences led at 36%. About 41% of dissertations by graduates educated in non-English-speaking countries had AI-flagged abstracts, versus 23% for those 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" for scientific careers, Daniel Gross, an economist on the paper, asked Nature.

Ang Liu, an engineer at the University of New South Wales in Sydney, told the magazine that when research pushes into areas of incomplete knowledge, "that is where rigorous research training becomes even more important." The study measures what Pangram flags in abstract text, not the quality of the underlying thesis, and the authors have not shown that AI use itself causes anyone to leave academia.