In 2026, artificial intelligence was responsible for generating text used in 30% of PhDs in STEM fields in the USA, as per a working paper analyzing dissertation abstracts submitted to an online repository since 2019.
The study, which has not yet been peer reviewed, also found that graduates who used AI when writing their dissertations are more likely to take industry jobs and less likely to stay in academia. As such practices grow, a broader question is being asked more often: where does tool assistance end and an essay writing service effectively begin, even when the role is played by a neural network rather than a person.
The researchers from the National Bureau of Economic Research (NBER) analyzed 226,000 dissertations in STEM obtained from U.S. universities in the period from January 2019 through May 2026. They used the AI detection service Pangram to assess the abstracts. Works were split into two categories: those with AI-generated text and those without. The study did not separate fully generated abstracts from ones only partly written with neural networks. According to co-author Daniel Gross, about 90% of the abstracts flagged as AI were classified by Pangram as entirely AI-generated.
AI-generated text first appeared in abstracts in 2023, when its share was 1.7%. By 2025 the figure had risen to 18.8%, and in the first five months of 2026 it reached 29.4%. Use was highest in mathematics and computer sciences, at about 36% of abstracts. Among graduates educated in non-English-speaking countries, 41% of abstracts contained AI-generated text, compared with 23% for graduates 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 Gross, an economist at NBER.
Ang Liu, an engineer at Australia’s University of New South Wales, says that generative AI works best in cases and fields that already have a large body of established knowledge. The trouble starts when research is done in areas where knowledge is still partial or unclear.
The authors also reviewed graduates’ LinkedIn profiles. Among those who completed their undergraduate degrees in the United States, graduates with AI-generated abstracts were 17% less likely to be employed in academia and 16% more likely to work in industry than those who did not use AI. For graduates who did their undergraduate studies outside the U.S., the link between AI use and academic employment was less pronounced.
