AI-generated text in US PhD dissertations rises sharply, with STEM and philosophy showing distinct trends

3 min read
Source: Daily Nous
AI-generated text in US PhD dissertations rises sharply, with STEM and philosophy showing distinct trends
Photo: Daily Nous
TL;DR

A new study indicates that AI-generated text is increasingly present in US PhD dissertations. In STEM fields, 29% of 2026 dissertations contain AI writing, up from zero in 2019. In philosophy, the rate rose from 1.3% in 2024 to 11.1% in 2026. The study links AI use to lower rates of academic career entry, particularly for US-origin students.

Key points

  • A working paper by Daniel P. Gross, Dror Shvadron, and Hansen Zhang analyzed US STEM PhD dissertations from 2019 to 2026.
  • AI-generated writing was absent before 2023 but reached 29% of STEM dissertations in 2026.
  • In philosophy, AI writing rose from 1.3% in 2024 to 7.8% in 2025, and 11.1% in 2026.
  • Students using AI were less likely to enter academic research careers and more likely to take non-tenure-track or industry roles.
  • AI use was more common among students from non-English-speaking countries and at lower-ranked programs.
  • The study used the Pangram API to detect AI writing, with a threshold of 2% AI-written text.

Background

The rise of AI in academic writing has prompted debates about the integrity of PhD dissertations. Previous concerns about AI use in undergraduate and graduate coursework have now extended to the final stage of doctoral training. The study builds on earlier analyses of AI use in academic writing and highlights the need for clear departmental policies.

How outlets are covering it

Daily Nous and Chemical & Engineering News both report the 29% figure for STEM dissertations in 2026, but Daily Nous provides additional detail on the philosophy data and the career outcomes of AI users. Daily Nous also highlights the methodological challenges in identifying philosophy dissertations, while Chemical & Engineering News focuses on the STEM findings. Some commentators in the Daily Nous comments section question the reliability of AI detectors, noting potential false positives and the difficulty of distinguishing between AI assistance and human writing.

Why it matters

The increasing use of AI in PhD dissertations raises questions about the integrity of academic credentials and the development of scientific expertise. If AI use becomes normalized, it could affect the expectations of administrators and the value of the PhD degree. Departments may need to implement policies to address AI use in dissertation writing, particularly as the trend grows.

What to watch

Departments and universities may need to develop or update policies regarding AI use in dissertation writing. Further research may be needed to understand the long-term effects of AI use on academic careers and the development of scientific expertise. The reliability of AI detectors may also be a focus of future studies.

Share this article

Want the full story? Read the original reporting

Read on Daily Nous