Using ChatGPT to generate a research dissertation and thesis. It is our research writing assistant.

ChatGPT can function as a “research writing assistant” that speeds up thesis and dissertation drafting—if the work is broken into small, well-scoped tasks and then recombined into a single document. The core workflow is to avoid dumping everything at once, instead feeding the model manageable chunks (abstract sections, analysis prompts, discussion, recommendations, and structure) and iterating until the output meets strict requirements like word limits.

The process begins with controlling length. After copying an abstract and checking its size (the example abstract is 340 words), the user prompts ChatGPT to “reduce the following to less than 250 words.” The first pass produces a shorter summary (129 words), but it also drops the original abstract’s subtitle structure. To preserve coverage, the workflow shifts to summarizing each subtitle individually, then combining those pieces. A second attempt yields a longer summary (141 words), and the user uses a “continue” command to extend the draft until it lands at an acceptable target (242 words total). The resulting summary is then saved and manually reviewed for small errors before being merged later.

Next comes turning results into thesis-ready narrative. The instructions emphasize that ChatGPT performs better when results are provided in organized paragraphs rather than tables, since the model may not reliably ingest tabular formats. The user pastes the results into ChatGPT with a prompt like “discuss the following results,” then uses “continue” whenever the response stops early. The output is copied into a separate document, reviewed, and later combined with the rest of the thesis materials.

From there, ChatGPT supports higher-level writing tasks: generating eye-catching research topics tied to the data, producing objectives and methodology (often separately, then combined), and drafting recommendations and conclusions from the discussion. The workflow also extends to research design decisions, including asking for a viable sample size (the example suggests 328) and requesting an outline for dissertation organization—especially useful for junior researchers who struggle with structure.

Finally, the transcript describes building citations using Semantic Scholar. Objectives are entered one by one into semantic scholar.org to surface relevant articles, which are then reviewed to select the best sources. The last step is compilation: merging all generated sections and articles into one document, then reading through to correct errors and improve flow. The takeaway is practical: chunk the input, format results as paragraphs, and iteratively refine by deleting or omitting content so the final thesis matches the desired outcome.