NBER: AI-written STEM dissertations tied to academia exit

Nearly 30% of US STEM dissertation abstracts filed by May this year contained AI-generated text, up from 1.7% in 2023, and the graduates whose work tripped the detector were measurably more likely to leave academia for industry. The finding comes from a working paper by Daniel P. Gross, Dror Shvadron and Hansen Zhang, surfaced on Nature by Mohana Basu; two researchers we follow had already shared the Nature piece.

The authors ran more than 226,000 STEM dissertations completed between January 2019 and May 2026 through Pangram, an AI-detection service. The share of flagged abstracts rose from 1.7% in 2023 to 18.8% in 2025, then 29.4% by May 2026. AI writing "is absent before 2023, present in 29% of dissertations filed in 2026, and rapidly growing," the NBER paper states.

Career outcomes diverge. Among US-origin graduates in the same program and cohort, those with AI-flagged abstracts were 17% less likely to pursue academic employment and 16% more likely to enter industry. Gross asks whether, "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" to go into a scientific career.

Usage is uneven. 41% of dissertations by graduates from non-English-speaking countries contained AI-generated abstracts, against 23% from English-speaking countries. The paper also finds AI use more common at lower-ranked programs, and highest in mathematical and computer science fields, where 36% of abstracts tripped the detector.

The flag is a signal, not a verdict. Pangram catches text that looks AI-generated; it cannot tell you whether a student drafted with the tool or merely polished with it.