Inside Higher Ed | Vladimir Zubkov/iStock/Getty Images Plus
For generations, the doctoral dissertation has been more than a scholarly document. It is a rite of passage that transforms students into scholars and is characterized by the defining quality of original thought. In dissertation defenses across the country, however, many faculty now grapple quietly with whether, and how much of, the work was produced with the help of generative AI.
This question is rarely asked of candidates themselves. Instead, faculty members and graduate programs are improvising, assuming that norms of authorship and originality will stretch far enough to accommodate these new tools. As students increasingly reach for generative AI to assist in analysis, synthesis and academic writing, the issue is no longer simply whether AI use threatens academic integrity, but whether the traditional dissertation still serves its intended purpose as the culminating expression of doctoral education in an age of AI-assisted knowledge production.
The dissertation’s authority rests on the assumption that its form is both timeless and necessary to demonstrate scholarly rigor. In reality, the modern dissertation reflects a particular set of conditions from when scholarship centered on individual authorship, linear argumentation and written text as the primary output of intellectual labor. These conventions reflected the dominant technologies of knowledge production at the time: the printing press and the typewriter, with the letter or telephone linking human collaborators. Over time, a printed dissertation became institutionalized as the standard measure of doctoral competence.
Yet the dissertation has always evolved alongside the tools and epistemic norms of its time. Few might remember the furor over word processing, which rendered the static nature of scholarly work fluid; more will recall the early days of the internet and concerns about plagiarism. These same technologies eased revision and collaboration, but the academic requirements for a dissertation remained the same: the doctoral student as the author of the text and the primary thinker behind the work. Generative AI–driven shifts continue this evolution, while challenging these standards more directly.
The current conversation about generative AI in doctoral education has largely focused on tools for writing, the use of which must be carefully monitored to prevent shortcuts or misconduct. This framing, while important, obscures a more fundamental academic shift. Generative AI does not simply assist with transcription or editing: It is increasingly integrated in the intellectual work of scholarship itself.
Scholars now use AI systems to explore and synthesize literature, scan for patterns in different types of datasets, evaluate alternative framings, suggest next steps and sequentially refine arguments. In these contexts, AI functions less as a passive instrument like the humble word processor and more as an epistemic partner. It is shaping how scholars pose questions, interpret evidence, shape ideas and communicate knowledge. Now that generative AI can be a part of the thinking process itself, the dissertation’s role as proof of independent intellectual achievement has begun to fracture.
The real disruption, therefore, is not that AI can polish student prose. Generative AI has instead blurred to the point of collapse long-standing boundaries around thinking, analyzing and composing. These are the very boundaries that structure standards for traditional academic work and doctoral dissertations. Doctoral training has historically treated dissertation writing as a culminating stage, documenting years of prior intellectual labor. Generative AI has fundamentally reframed what was once a linear process of thinking into an iterative process of exploration that may no longer be the predominant effort of a single individual. Synthesis becomes dialogic and cannot be evaluated using the current markers of independence. Such changes challenge the assumption that intellectual autonomy is best demonstrated by solitary authorship, raising fundamental questions about how doctoral work is mentored and evaluated.
Some have taken a strict approach, requiring complete abstinence from generative AI in dissertation research and writing. Others are addressing these changes more flexibly, often on a case-by-case basis. At a time when the promise and challenge of AI integration is unclear, programs are exploring new ways to support rigorous scholarship. However, striking a balance between innovation and clear standards for legitimate work remains difficult. Graduate schools across the country are updating guidance and ethics statements, yet these policies rarely tackle how AI reshapes the process of thinking or constructing arguments. A patchwork of campus policies attempts to treat AI as a compliance issue, but doctoral education must instead confront how the dissertation, and academic research writ large, is evolving.
Generative AI tools can transform research in fields that center humanism. Scholars in the humanities have long relished their time in archives, an undertaking that has already been altered by the digitization of many archival collections. AI promises to speed the process of finding examples that used to take hours, days or even weeks to identify. Consider a scholar who publishes work that requires sifting through early-20th-century periodicals to find essays, images and classified ads that highlight particular objects. Would using a large language model that could quickly identify potential examples limit the value of that research? The answer is likely no: The interpretive work once one identifies examples remains unchanged. Moreover, generative AI promises to expand archival access to doctoral students, who usually lack the resources to travel to distant libraries in expensive locations.
The integration of generative AI could, additionally, reshape how dissertations are conceived and executed in STEM fields. AI enables the rapid exploration of complex patterns across massive datasets in ways that were previously unimaginable. Rather than simply reporting results after manual analysis, students can engage in an ongoing dialogue with AI systems that test hypotheses against these datasets and refine their questions in real time. This changes the dissertation from a static report of prior work into a dynamic artifact. The dissertation in these disciplines may now reflect a record of discovery that includes human-AI collaboration as an explicit part of the intellectual process.
In this way, dissertations could become not just evidence of independent thinking, but documentation of how novel insights emerge at the interface of human and machine intelligence. Imagine research aimed at identifying patterns within cell morphology by fitting cells into common, regular shapes. Would using an LLM that could identify alternative patterns have limited the value of this research? The answer is no: The value may actually increase by finding morphologies that were not encoded into customized screening tools.
As doctoral programs explore new possibilities, institutions must rethink the structures that support graduate scholarship. Mentorship can no longer focus on what happens in the lab or library; it must also guide students in the iterative, human-AI scholarship processes that can generate new knowledge by modeling effective AI engagement. Students will need help in framing questions, evaluating outputs critically and reflecting on how machine partners shape thinking.
By focusing on how the process of research has evolved in addition to the dissertation as a culminating product, doctoral education can cultivate scholars who are not only technically proficient but also intellectually agile in navigating AI-mediated environments. Assessment and evaluation of the dissertation will also require reimagining. Traditional markers of independence and intellectual maturity will become insufficient when AI aids in student research, and new forms of documentation and citation may need to be developed. Such steps will help to preserve rigor and accountability while recognizing that originality and scholarly contributions can emerge from collaborative human-machine processes.
Embracing AI as a partner invites a broader conception of the dissertation itself. Across disciplines, the culminating work could become a more dynamic, process-oriented artifact, one that integrates experimental and intellectual iteration with AI-assisted analysis. Such a model does not diminish scholarly achievement but instead expands it, aligning doctoral education with the realities of contemporary knowledge production.
The challenge for universities is to adapt expectations, mentoring practices and assessments to this new reality, ensuring that doctoral training remains meaningful and relevant. Without adapting our ideas of authorship and assessment, doctoral education risks preserving the appearance of rigor while quietly abandoning its very foundations.
