NBER: AI-using PhDs 17% less likely to enter academia

Thirty percent of STEM PhD dissertation abstracts filed in the US this year contained AI-generated text, according to a National Bureau of Economic Research working paper reported by Nature. The share was 1.7% in 2023 and 18.8% in 2025. Among mathematical and computer science abstracts, 36% showed AI writing.

Researchers scanned more than 226,000 abstracts from US STEM doctoral graduates completed between January 2019 and May 2026, using the AI-detection service Pangram. About 90% of flagged abstracts were classified as entirely AI-generated rather than partially rewritten.

The employment correlation is the paper's sharper claim. Graduates whose abstracts tested positive for AI were 17% less likely to work in academia and 16% more likely to go into industry. The split by background is also stark: 41% of dissertations by graduates from non-English-speaking countries had AI-generated abstracts, versus 23% 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 to be good scientists?" asks Daniel Gross, an economist at NBER in Durham. Ang Liu, an engineer at the University of New South Wales in Sydney, says "rigorous research training becomes even more important" when research enters areas with incomplete knowledge.

The working paper has not been peer-reviewed, and the study did not generally distinguish between abstracts partly rewritten with AI and those drafted end-to-end. Two of the scholars we track in Who's Who circulated the Nature write-up the day it ran.