NBER paper links AI-written PhD abstracts to industry exits

Nearly 30% of US STEM PhD dissertation abstracts filed in the first five months of 2026 contained AI-generated text, up from 1.7% in 2023, according to a working paper covered by Nature.

The authors, Daniel P. Gross of Duke, Dror Shvadron of Toronto and Hansen Zhang of Duke, scanned more than 226,000 STEM dissertations completed between January 2019 and May 2026 with the Pangram AI-detection service. "AI writing is absent before 2023, present in 29% of dissertations filed in 2026, and rapidly growing," they write in NBER Working Paper 35859.

Adoption splits sharply along two lines. 41% of graduates from non-English-speaking countries had AI-flagged abstracts, against 23% from English-speaking countries, and use was also more common at lower-ranked programs.

The career correlation is what gives the paper its edge. Among US-educated PhDs, those with AI-flagged abstracts were 17% less likely to pursue academic careers and 16% more likely to work in industry than peers without the flag. The authors frame it as a trade: "AI may increase research productivity, it thus appears to also be displacing a potentially important portion of the work through which scientific expertise is developed."

The paper does not say whether AI-flagged graduates are opting out of the tenure track or being filtered out of it. The Pangram label is a probabilistic signal, not a confession. Philosophy, which the authors look at separately, is tracking the STEM curve on a lag: 1.3% of dissertations in 2024, 7.8% in 2025, and 11.1% in early 2026 samples.