NBER: 29% of 2026 US STEM PhD dissertations contain AI writing

Twenty-nine percent of US STEM PhD dissertations filed in 2026 contain AI-generated writing, up from zero before 2023, according to a new NBER working paper by Daniel P. Gross of Duke, Dror Shvadron of Toronto, and Hansen Zhang of Duke.

The authors measured AI-generated content across a near-census of US STEM PhD graduates from 2019 to 2026. "AI writing is absent before 2023, present in 29% of dissertations filed in 2026, and rapidly growing," the paper reports. Use is more common among students from non-English-speaking countries and at lower-ranked programs.

The career data is where the paper lands its sharpest claim. Among graduates in the same program and cohort, those whose dissertations contain AI writing are less likely to enter academic research careers and more likely to end up in non-tenure-track roles or industry. The pattern is strongest among US-origin students.

The authors frame the stakes plainly: "delegating research writing to AI may undermine the development of expertise that scientific training is supposed to build." They concede AI "may increase research productivity" but argue it "appears to also be displacing a potentially important portion of the work through which scientific expertise is developed."

Two researchers we track shared the paper the week it went up. The abstract does not publish detector accuracy numbers, breakdowns by STEM discipline, or which sections of the dissertation carry the heaviest AI load.