AI-generated writing is increasingly present in US STEM PhD dissertations, rising from near-absence before 2023 to 29% in 2026, with philosophy dissertations showing a smaller but rising trend (about 1.3% in 2024, 7.8% in 2025 and 11.1% of 2026 projections analyzed so far).

Overall, the study highlights both the prevalence of AI-generated content in dissertations and potential career implications, suggesting a shift in academic versus industry pathways linked to AI-assisted scholarly writing.

There is a substantial rise in the use of generative AI tools for scientific writing across science, technology, engineering, and math fields within PhD dissertations.

The forward-looking takeaway is that departments should proactively develop policies and norms to address AI use before it becomes fully normalized, with room for input and debate.

Experts caution that relying on AI for writing can obscure rigorous research in areas with incomplete or uncertain knowledge, underscoring the need for robust training in traditional research methods.

Job outcomes show that US-undergraduate graduates with AI-generated abstracts are 17% more likely to enter industry and 16% less likely to stay in academia, with variations for those trained outside the US.

Limitations include challenges in identifying philosophy dissertations due to metadata issues and evolving AI use, raising concerns about misclassification and the precise definition of AI-written content.

In STEM, AI-writing is more prevalent among non-English-speaking students and lower-ranked programs, with those having AI-written dissertations tending toward non-tenure-track roles or industry rather than academic research.

A 2024–2025 working paper found that PhD graduates who used AI to help write their dissertations were more likely to work in industry and less likely to enter academia.

In philosophy, AI writing shows a parallel upward trend (at least 2% AI-written text), suggesting discipline-wide normalization risks similar to STEM, albeit at a smaller scale.

Policy interpretations center on how to craft fine-grained, enforceable AI-use guidelines for dissertations and whether norms will shift as AI use becomes more widespread.

Daniel Gross of Duke University notes the trend will likely continue as generative AI becomes more integrated into scientific writing practices.