NBER: AI-written PhD dissertations tied to industry careers

Among US-origin PhDs whose dissertation abstracts tested positive for AI writing, the group was 17% less likely to enter academic employment and 16% more likely to move into industry than classmates in the same program and cohort, Nature reported on 16 October, citing a new NBER working paper by Daniel P. Gross, Dror Shvadron and Hansen Zhang.

The scale is recent. "AI writing is absent before 2023, present in 29% of dissertations filed in 2026, and rapidly growing," the paper reports. Prevalence of AI-flagged abstracts climbed from 1.7% in 2023 to 18.8% in 2025 before reaching 29.4% in early 2026. The authors examined more than 226,000 STEM dissertations completed between January 2019 and May 2026 at US universities, flagging abstracts via the Pangram AI-detection service.

Use is uneven. 41% of dissertations from graduates of non-English-speaking countries carried AI-written abstracts, against 23% from English-speaking countries, with heavier use at lower-ranked programs. The academia-to-industry gap, by contrast, was strongest among US-origin students, where background and language cannot be invoked to explain the pattern.

The authors frame AI's role as a trade-off: it may lift research productivity, but it also displaces the writing-heavy developmental work through which scientific expertise typically forms during doctoral training.