From the Research Universities edition of October 16, 2026
NBER working paper: AI writing appears in 29% of 2026 US STEM PhD dissertations, and its users are less likely to enter academic research
National Bureau of Economic Research (Daniel P. Gross, Dror Shvadron, Hansen Zhang); reported by Nature 16 Oct 2026 · Doctoral research training and the research workforce · United States (US STEM PhD graduates 2019-2026)
- Publisher
- National Bureau of Economic Research
- Original publication
- NBER Working Paper 35859, issue date October 2026 (day not stated); Nature news coverage 16 October 2026
- Source retrieved
- 2026-10-17
What happened
NBER Working Paper 35859 (Gross, Shvadron and Zhang; issue date October 2026) measures AI-generated writing in a near-census of US STEM PhD dissertations from 2019 to 2026 and links graduates to later employment. AI writing is absent before 2023, present in 29% of dissertations filed in 2026, and rising; use is more common among students from non-English-speaking countries and at lower-ranked programmes. Within the same programme and cohort, graduates whose dissertations contain AI writing are less likely to enter academic research careers and more likely to enter non-tenure-track roles or industry, strongest for US-origin students. The authors argue AI may be displacing part of the work through which scientific expertise is developed. Nature's 16 October report on the paper adds that it covers more than 226,000 graduates, used the Pangram detector on abstracts only without separating partial from full AI generation, found 1.7% in 2023, 18.8% in 2025 and 29.4% in January-May 2026 (36% in mathematical and computer sciences), and that US-undergraduate graduates with AI-written abstracts were 17% less likely to work in academia and 16% more likely to work in industry; an engineer at UNSW warns training matters most where knowledge is incomplete.
Why it matters
Graduate schools and research offices set dissertation standards and train the next research workforce. A near-census finding that 29% of 2026 STEM dissertations contain AI writing, with an association to leaving academic research, gives them a measured baseline for policy on AI use in doctoral work and for reviewing whether training still builds expertise.
Evidence and measured results
Near-census of US STEM dissertations 2019-2026 (Nature: >226,000 graduates), detector applied to abstracts, linked employment profiles; within program-cohort comparisons; working paper.
Limitations and uncertainty
Not peer reviewed; detector-based; abstracts only; correlational career outcomes; LinkedIn coverage bias.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-16; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Graduate schools lack a measured baseline and policy for AI use in dissertations and worry training no longer builds research expertise.
- Stakeholders
- Graduate school dean
- Vice president for research
- Faculty advisors and committee chairs
- Office of international students
- Research integrity office
- Discovery questions
- Does the graduate school have a dissertation AI-use and disclosure policy?
- How do committees assess how research and writing were done?
- How would policy affect students writing in a second language?
- Value hypothesis
- Interpretation: clear disclosure and advising practices may protect training quality without banning useful AI assistance.
- Potential engagement
- A policy and practice review for dissertation AI use, with draft disclosure language and advisor guidance.
- Unsupported claims
- That AI use causes students to leave academia
- That detector scores identify individual misconduct
- Peer-reviewed status or findings outside US STEM
Pre-sales engineering
Role takeaway
- Fit
- Limited; relevant only if institutions measure or screen thesis text.
- Architecture and integration
- Aggregate, de-identified analysis of thesis abstracts with a validated detector, separated from individual decisions.
- Prerequisites
- Consent and data-use approval
- Labelled validation sample
- Detector with published error rates
- Constraints
- Detector error rates and bias
- Abstracts may not represent full documents
- Legal and privacy limits on student work
- Security
- Minimize data shared with vendors, set retention limits and avoid individual-level scoring.
- Proof of value
- Validate detector on a labelled local sample
- Report only aggregate prevalence
- Review results for language-background bias
Delivery
Role takeaway
- Work
- Draft dissertation AI-use and disclosure policy, train advisors, and update committee review practice.
- Dependencies
- Faculty governance approval
- Graduate school systems
- Student input
- Ownership
- Graduate school dean with faculty council.
- Skills and adoption
- Advisor workshops on supervising AI-assisted research and writing; student guidance on disclosure.
- Governance checkpoints
- Faculty council approval
- Equity review for non-native English speakers
- Annual review of policy
- Acceptance criteria
- Proposal: disclosure statement in all dissertations filed after adoption
- Proposal: advisor training completion tracked
- Proposal: no individual sanctions based on detector scores alone
- Risks
- Penalizing non-native speakers
- False accusations
- Policy that ignores legitimate AI use
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Limited; institutions considering AI-detection in thesis workflows should treat detector output as a screening signal with known error rates, not proof.
Governance
Who approves, reviews and stays accountable for outcomes?
Graduate schools may need explicit dissertation AI-use and disclosure policies and advisor guidance; the paper measures prevalence, not misconduct.
Security and privacy
What data, permissions and controls need testing?
Running student work through commercial detectors raises data-handling and consent questions.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Higher use among students from non-English-speaking countries suggests AI is also a language aid; policy should avoid penalizing those students while protecting training quality.
Procurement
What should contracts, pricing and exit terms secure?
Any detector procurement should require published error rates and data-retention terms.
Operating model
Which teams own the service once it runs?
Advising and committee review may need to check how research and writing were done, not only the final document.
Publication history
- 2026-10-16Research Universities · Issue 385 resources
Stable resource ID: nber-w35859-ai-writing-us-stem-phd-dissertations-careers-202610