AI Use in STEM Dissertations Surged by 2026

Nearly 30 percent of STEM doctoral candidates used AI for writing by mid-2026, up from less than 2 percent in 2023.

Updated on Oct. 5, 2026 in Artificial Intelligence

Isometric editorial illustration showing a stack of hardbound research volumes on a plain, minimalist surface. — Nearly 30 percent of STEM doctoral dissertations filed in the United States by mid-2026 contained AI-generated material, up from less than 2 percent in 2023. AI Illustration. Upload story photo >

Live Poll

Do you believe using AI to write academic work hinders the development of true professional expertise?

A working paper documented that 29.4 percent of STEM dissertations filed in the United States between January and May 2026 contained AI-generated material. This figure represents a significant increase from 1.7 percent in 2023, following the public release of ChatGPT in the fall of 2022.

Why it matters

The rapid adoption of AI tools in doctoral research highlights a fundamental shift in academic writing practices across technical fields. Understanding this trend is critical for evaluating how language model integration influences scientific communication and the integrity of advanced academic training.

Researchers utilized Pangram, an AI detection tool, to screen dissertations against content produced since the public rollout of ChatGPT. Math and computer science fields led usage with 36 percent of abstracts containing AI content, compared to 22 percent in the physical sciences.

The players

Duke University

An academic institution serving as the home base for the study's lead investigator.

Massachusetts Institute of Technology

A research university that hosted specific committees investigating AI adoption in academic writing.

The details

The analysis compared the density of machine-generated text in dissertations against the timeline of AI tool availability. Non-native English speakers utilized AI tools more intensively than their native-speaking peers during the research and writing process. The Pangram tool classifies writing as AI-generated by detecting patterns typical of large language models—algorithms trained on massive datasets to predict subsequent text tokens.

Fall 2022: ChatGPT became widely available to the public.
2023: 1.7 percent of dissertations included AI-generated material.
2024: 5.5 percent of dissertations included AI-generated material.
2025: 18.8 percent of dissertations included AI-generated material.
January to May 2026: 29.4 percent of dissertations included AI-generated material.

The Tech Race

The adoption of large language models in higher education has accelerated rapidly since the public rollout of ChatGPT. This trend indicates that academic institutions are following a similar integration trajectory to software and professional services sectors currently racing to implement generative tools.

This development indicates that technical research workflows now frequently rely on AI-assisted tools for drafting and structuring content. Students and faculty should anticipate updated institutional policies regarding the transparency and permitted use of language models in future dissertations.

The takeaway

The data confirms that the use of generative AI in high-level scientific writing has become a standard practice rather than an outlier. Readers should watch for forthcoming academic policies and updated verification standards from university research boards throughout the 2026-2027 school year.