Video summary

The AI Thesis Writing System That Cuts Your Workload in Half

Main summary

Key takeaways

Educational

Main ideas / lessons conveyed

  • Writing a thesis can feel overwhelming, but AI can substantially reduce workload by helping you structure content, draft text from figures, generate/organize the introduction from sources, and polish academic quality.
  • AI tools still require real research outputs (“data”)—they won’t generate your primary findings from nothing.
  • A step-wise workflow works best: figures → story structure → figure text → (later) introduction → remaining sections → academic editing.

Methodology / workflow (detailed step-by-step)

1) Start with data you already have (required foundation)

  • Recognize that AI won’t create thesis “data” (at least currently).
  • Ensure you have many research results / materials you need to report and discuss.
  • While doing research (e.g., PhD), collect “research stories” as you go.

2) Organize by chapter, then plan figure volume

  • Prefer organizing/thinking by chapter rather than by a whole thesis outline.
  • Estimate figure volume based on your thesis style:
    • Example guideline: ~12–15 figures per chapter.
    • Adjust depending on field norms (some fields may expect more or fewer).
  • Treat each chapter as a set of figures that must become a coherent narrative supported by citations.

3) Use AI vision to generate an outline/story structure from figures

  • Upload figures to ChatGPT (AI vision).

  • Prompt example (conceptual):

    “These are figures forming a chapter in my thesis—suggest a story structure for the outline.”

  • Expected output includes:

    • A chapter-level structure (e.g., introduction + logical sections)
    • A sense of flow implied by the figure sequence
    • Section ideas such as:
      • Fabrication/process descriptions
      • Electrical properties / other domain-specific headings

  • Goal: use the figures as the “destination,” letting AI propose the logical structure that matches your data.

4) Draft text by building paragraph-by-paragraph from each figure (voice-to-text)

  • After obtaining story structure, draft the chapter content.

  • Workflow:

    • Switch to phone
    • Use a voice input feature to dictate what you know about each figure

  • For each figure:

    • Tell AI details casually (what’s in each panel, what changes, what results show).
    • Example described:
      • A figure with four panels (A, B, C, D)
      • AI outputs an initial drafted paragraph
    • Then use a prompt like:
      • “Convert this into a paragraph for my thesis.”

  • Result: you get text directly under/around each figure that builds the narrative from:

    • structure → figures → text

5) Write the introduction after results/discussion text exists (conceptual suggestion)

  • The introduction is described as very important because it provides:
    • background / foundational knowledge
    • concepts needed to understand the thesis

6) Generate introductions using literature sources (two-tool approach)

Option A: NotebookLM (outline-first approach)

  • Use NotebookLM with a set of papers (example: “20-something papers”).

  • Prompt approach:

    “Write an introduction section to my thesis based on these sources.”

  • If output is too short, ask for an expandable outline, e.g.:

    “Create an outline… that could be expanded into 20,000 words.”

  • Benefit:

    • Produces a more useful structured outline
    • Includes reference support so claims are grounded

Option B: ThesisAI (directly produces long, referenced draft)

  • Use thesisAI.
  • Input: up to 100 reference papers
  • Output: up to ~17 pages of intro-like material
  • Features:
    • Fully referenced
    • Includes an abstract
    • Can be exported/copied into Overleaf with control over citations/layout
  • Caveat noted:
    • Not guaranteed; “your mileage may vary.”

7) If you don’t have enough references: use deep research

  • Use Perplexity AI (described as best at the time for this task).

  • Use “deep research” mode:

    “Research [topic] for the introduction section… Write the introduction section based on the research.”

  • Output:

    • provides a significant amount of research/data in a usable format
  • Then:
    • take the references produced
    • import them into a citation manager
    • expand/build the introduction from there
  • Goal:
    • create a strong starting point even when you’re missing key sources.

8) Draft remaining thesis sections using “combination” or communication methods

  • Conclusions section:

    • Take conclusions from each chapter
    • Paste them into an AI model and ask for:
      • a combined “mega conclusions” paragraph/section
    • Instruct AI to reference:
      • “these are conclusions from Chapter 1/2/3… combine into the end-of-thesis conclusion.”

  • Alternative described:

    • use voice-to-AI again to draft sections quickly (“talk into your phone”).

9) Methods section: AI is “least useful,” but can help generate a record-based draft

  • Methods are largely factual and must reflect what you truly did:
    • lab activities
    • instruments used
    • products purchased
    • techniques used
  • AI can help if you provide the details:

    “Here are the methods I used. Write a methods section for my thesis.”

10) Ensure academic standards with AI editing tools

Two editing/polishing tools are recommended:

Tool 1: PaperPal

  • Works alongside writing to provide:
    • suggestions
    • checks to keep content within academic standards
  • Integrations noted:
    • Microsoft Word
    • Google Docs
    • Overleaf

Tool 2 (preferred): Thesify

  • Described as doing heavier lifting than a reluctant supervisor.
  • Input:
    • peer-reviewed paper, thesis chapter, or document
  • Outputs described:
    • feedback summary (what works, what to improve)
    • recommendations
    • ability to drill into areas like purpose (and other aspects)
    • feedback framed like a helpful supervisor
  • Recommended use:
    • run your draft through Thesify before sending to your supervisor.

Sources / speakers featured (as mentioned)

  • Speaker/host (implied): “Andy” is mentioned (likely the creator/host addressing another person). No other named speakers are clearly identified.
  • AI tools / platforms referenced:
    • ChatGPT (AI vision; voice-to-text)
    • NotebookLM
    • thesisAI (spelled “thesis AI” / “Thesis AI” in subtitles as “thesis AI”)
    • Perplexity AI
    • Claude (mentioned as an alternative AI model for methods/sections)
    • PaperPal
    • Thesify
  • Other services mentioned:
    • Google Research Co-pilot (future warning)
    • Overleaf
    • Citation manager (generic; no specific product named)

Original video