PhD students using AI in dissertations less likely to enter academia
A working paper reviewed 226,000 STEM doctoral dissertations completed between January 2019 and May 2026 in the United States. Roughly 30 % of these dissertations contained AI‑generated text, according to the study. AI use first appeared in 2023, rising from 1.7 % that year to 18.8 % in 2025 and reaching 29.4 % in the first five months of 2026. The practice was most common in mathematics and…
Key points
- 30 % of STEM dissertations in the US contained AI‑generated text this year
- AI use in abstracts rose from 1.7 % in 2023 to 18.8 % in 2025, reaching 29.4 % in early 2026
- Graduates with AI‑generated abstracts were 17 % less likely to work in academia and 16 % more likely in industry
The paper also examined career outcomes of graduates. Those with AI‑generated abstracts were 17 % less likely to be employed in academia and 16 % more likely to work in industry compared with peers who did not use AI. The effect was weaker for graduates who completed their undergraduate studies outside the United States. Researchers from the National Bureau of Economic Research raised questions about whether heavy reliance on AI compromises the depth of scientific understanding.
The study used the Pangram AI‑detection service, which flagged 90 % of the AI‑generated abstracts as entirely AI‑written. The paper has not yet been peer reviewed.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- KDnuggets lists five resources to learn about world models in AI · 1 src
- Llama.cpp merges AVX2 tiled accumulator reduction update · 1 src
- GLM-5.3 shows rare update-aware reasoning during training · 1 src
- OpenAI offers U.S. university labs free ChatGPT Business seats with Rosalind access · 2 src
- Anthropic reports small AI models may self-replicate on IoT devices within months · 29 src
Comments
via GitHub Discussions