General educational commentary: This article is not legal, career, or institutional advice. Employment outcomes vary. Follow your institution's policies and consult appropriate advisors.

The question is no longer whether AI has entered doctoral writing. New research suggests it already has. The more important question is what doctoral writing demonstrates about the researcher behind it.

Strictly speaking, the study measures detector-flagged AI writing in dissertation abstracts. It does not measure self-reported AI use.

An October 2026 NBER working paper — a non-peer-reviewed preprint — reports a sharp rise in AI-generated text in U.S. STEM dissertation abstracts. Its findings are part of a related debate in The Chronicle of Higher Education: Would You Hire a Ph.D. Who Wrote Their Dissertation With AI? (subscription required).

The answer is not a simple yes or no. Responsible use of AI does not automatically disqualify a graduate. But a dissertation is more than a credential. It is evidence of how a scholar thinks, makes judgments, and communicates expertise. If AI takes over the work that develops those abilities, the cost may reach beyond the dissertation.

What the new research found, and what it did not

The NBER authors examined a near-census of U.S. STEM doctoral dissertations filed from 2019 through May 2026. They used a commercial AI-writing detection service (Pangram) to evaluate dissertation abstracts, then tested how well those results corresponded with AI writing in a smaller sample of full dissertations.

Two clarifications about the method. First, the authors describe their results as suggestive rather than definitive. Second, the paper's acknowledgements note that the detection vendor provided API credits and discounted pricing. Panther has no relationship with that vendor. Readers should weigh detector-derived findings alongside the concerns about detection accuracy discussed elsewhere in this blog.

The share of abstracts flagged as containing AI-generated writing rose from 1.7% in 2023 to 5.5% in 2024, 18.8% in 2025, and 29.4% in the 2026 cohort. The paper estimates that about 31% of dissertations in the 2026 cohort may contain AI writing in their full text. These figures are not a count of dissertations written entirely by AI. Nor do they establish how much of any one student’s work was delegated.

The researchers also examined graduates’ employment profiles. Among U.S.-origin graduates from research-intensive universities, AI writing in a dissertation abstract was associated with a lower likelihood of academic employment and a higher likelihood of industry employment. The authors are careful to describe this relationship as correlational, not causal. It does not show that employers rejected candidates because of AI use. It does not establish what hiring committees think.

That distinction matters. The study raises a serious question about doctoral training. It does not settle the question of whether an individual graduate is qualified or employable. A Chemical & Engineering News report on the research likewise notes both the rapid adoption and the uncertainty about its long-term effects.

A dissertation is also evidence of how you think

A dissertation is a substantial investment of time, tuition, and professional effort. Its value is not only the document it produces. The process requires a doctoral student to interpret evidence, build an argument, make research decisions, and explain why those decisions are defensible.

Writing is part of that work. A student who has to explain why studies disagree may discover a weakness in the argument. A student who describes a method precisely may recognize an assumption that needs attention. Revising a chapter can reveal a gap between what the data show and what the conclusion claims.

That is why academic voice matters. It is not a matter of sounding grand or using complex vocabulary. It is the identifiable quality of a researcher’s reasoning: the claims they make, the evidence they prioritize, the limits they acknowledge, and the way they connect their findings to the field.

A dissertation can be polished and still lack a clear scholarly position. Conversely, a writer can have strong ideas that are obscured by dense sentences or inconsistent structure. The goal is not to make every sentence sound like an expert’s. It is to ensure that the scholar’s own analysis is clear enough to be assessed.

Would a committee or employer hire that PhD?

Some would. Some may ask closer questions. The NBER study does not provide evidence that hiring committees are screening candidates based on AI use. Its employment findings should not be treated as proof of a hiring penalty.

Still, employers and academic committees have good reasons to assess the skills a dissertation is supposed to develop. Can the candidate explain the research without relying on a prepared paragraph? Can they defend their methodological choices? Can they identify the limitations of their findings and respond thoughtfully to challenges?

For academic roles, those abilities are directly relevant to research, teaching, grant applications, and publication. Outside academia, employers also value independent judgment, clear communication, and the ability to evaluate complex information. A degree is a signal of training. The candidate’s command of their work makes that signal credible.

The central risk, then, is not that a dissertation contains any AI-assisted text. It is that the graduate cannot demonstrate the reasoning the dissertation is meant to represent. A candidate who has used tools thoughtfully and can explain every claim may present a stronger case than one who relied on polished language without developing a firm grasp of the argument.

How to build an academic voice that remains yours

Doctoral students can use technology without surrendering the intellectual work. The practical test is simple: keep ownership of the reasoning, not just the final document.

• Draft your central claim before asking for outside input. State what you believe your research shows and why. Then compare that claim against your evidence.
• Synthesize sources yourself. Identify where studies agree, where they conflict, and what explains the difference. A list of summaries is not a scholarly argument.
• Write down your research decisions. Explain why you selected a design, measure, or analytical approach. This record helps you prepare for committee questions and recognize the logic behind your choices.
• Use feedback to sharpen your thinking. Tools or colleagues can point out unclear passages, gaps, or possible counterarguments. You should decide how to respond and revise in your own judgment.
• Practice explaining your work aloud. If you cannot describe a key paragraph, finding, or limitation in plain language, revisit it before submission.

These habits take effort. They also deliver a meaningful return. They prepare you for your defense, strengthen your future research and professional communication, and help you make the most of the years already invested in your degree.

Editing can improve expression without replacing expertise

Strong academic voice does not require a student to work alone. An editor can help identify unclear sentences, tighten structure, improve consistency, and apply a required style. That support can make a researcher’s argument easier to follow. It should not replace the researcher’s analysis or take ownership of the scholarly claims.

That distinction is central to ethical, effective academic editing services. Professional dissertation editing services can support clarity and presentation while leaving the research decisions and final judgment with the author. You can review Panther's document submission process before you begin. Coaching can also help students work through the reasoning and writing challenges behind a difficult chapter.

For doctoral students, this can be a cost-effective way to protect time and improve the manuscript without outsourcing the milestone that matters most: demonstrating command of the work. Panther Academic Editing provides academic editing and dissertation coaching for students and researchers. Its editors specialize in academic documents and work with styles including APA, Harvard, Chicago, and Turabian.

The milestone employers will remember

The NBER findings deserve attention, but they do not justify a blanket verdict on PhDs who use AI. The study focuses on U.S. STEM dissertation writing and reports associations, not proof that AI use causes a particular career outcome or that employers reject AI users.

The stronger conclusion is practical. Your dissertation is both a research product and evidence of your development as a scholar. Protect the work that builds your ability to make claims, weigh evidence, explain decisions, and answer for your conclusions. Use support to make that work clearer, not to conceal gaps in understanding.

A future employer may never ask which tools you used. They may ask you to explain what you learned, how you reached a conclusion, and what you would do next. Your academic voice is the answer. Make sure it is unmistakably yours.