AI Tools for Enhanced Thesis Writing

AI tools can turn thesis writing from a slow, solitary struggle into a faster workflow—by helping draft more academic text, supplying discipline-specific terminology, and accelerating the search for credible references and evidence. The biggest practical shift is using large language models not just for “writing,” but as a configurable research assistant that matches a student’s context and desired academic tone.

A key starting point is customizing ChatGPT. Instead of treating it like a generic chatbot, the workflow recommends clicking into “customize chat GPT” and adding custom instructions: the user’s role (thesis writer), the level of formality needed (academically focused language), response length (longer answers to generate more material for editing), and a strict preference for neutrality (no opinions, fact-based and science/data-driven output). With those settings, the model can help rewrite paragraphs to sound more thesis-ready—such as making a paragraph more formal and adding figure-style academic phrasing.

That said, the transcript draws a clear boundary: AI can’t invent results. For thesis claims tied to experiments, the student must provide the actual findings (e.g., what an AFM dataset shows). Once the results are supplied—often by describing a figure or embedding figure-specific information—AI can help shape the narrative around them, including more academic wording and structured descriptions that the student can then accept, cut, or revise.

The second major use case is terminology mastery. When writing about specialized processes, students often need more than plain-language descriptions. Prompts like “best way to describe the process of delamination” can generate discipline-appropriate wording and a more complete causal chain. In the delamination example, the model frames delamination as layers separating in composite materials and suggests related concepts such as stress concentration, fatigue, imperfections, and fatigue propagation—useful for describing how transparent electrodes fail or degrade. The result is writing that sounds professional and technically precise rather than “high school level.”

Finding the “perfect reference” is treated as another bottleneck that AI can ease. Tools such as “scite assistant” (site. a) and “s iace” are mentioned as AI research partners for locating citations, generating referenced answers, and extracting or paraphrasing information from specific PDFs. The workflow emphasizes setting response preferences—like reference structure and recency—so the output aligns with thesis expectations.

Finally, AI proofreaders help prevent supervisors from getting stuck on minor mechanics. Trinka is highlighted for producing a tracked-changes Word document with comments, while Paperpal is described as offering side-by-side checks after pasting sections (including issues flagged and suggested edits). The underlying goal is to reduce distraction from punctuation and formatting so feedback can focus on the science itself—what the research actually shows and how convincingly it’s argued.