An honest comparison

ChatGPT for thesis writing — what it does well, and where it stops

A general chatbot is a genuinely good thinking partner. It is not a citation system. This page sets out the difference plainly, so you can decide which part of the work belongs to which tool.

No card required · every citation traceable to a real source

Three differences that decide a thesis

Where the answer comes from

A chatbot answers from what the model holds plus whatever it retrieved for that message. A source-grounded workspace answers from a library you built and can open, so a reference is a publication record with metadata, not a plausible-looking string.

What happens when the evidence is thin

Asked for support it does not have, a language model will often produce something that reads like a citation. Grongy AI leaves the sentence uncited instead — a worse-looking draft, and a better one to hand in.

What survives between sessions

A chat is a conversation; a thesis is a document. The outline, the sources, the notes and the decisions have to still be there next week, and they have to shape what gets written next.

How it works in practice

The useful answer is not one tool or the other. Most people who finish a thesis end up using a chatbot to think and a document workspace to write.

  1. 1

    Use a chatbot to think

    Sharpen the research question, argue against your own reasoning, get a plain-language explanation of a method you have not used before.

  2. 2

    Bring the sources into a library

    Search scholarly indexes, save the papers that belong in the project, upload the PDFs you already have — and read them.

  3. 3

    Draft from the library, not from memory

    Write each section against the papers you saved, so every citation resolves to a record you can open and check.

  4. 4

    Verify, rewrite, disclose

    Check the claims, put the prose in your own voice, and state your AI use wherever your institution requires it.

Side by side

Written about general-purpose chatbots as a category, not about one product's current feature list — those change every few months, and a page that pinned itself to them would be wrong by next term.

  1. 1

    Where an answer comes from

    Chatbot: the model, plus whatever it retrieved for that message. Grongy AI: the papers saved in your project library.

  2. 2

    References

    Chatbot: text shaped like a reference, which you check one by one. Grongy AI: publication records with metadata and a link to open.

  3. 3

    Unsupported claims

    Chatbot: usually smoothed over in fluent prose. Grongy AI: left visibly uncited, so you can see what still needs evidence.

  4. 4

    Project memory

    Chatbot: whatever fits in the conversation. Grongy AI: outline, sources, notes and key decisions kept with the document.

  5. 5

    The document itself

    Chatbot: text you copy out. Grongy AI: headings, table of contents, footnotes and a reference list, exported to Word, PDF, LaTeX or Markdown.

  6. 6

    Citation styles

    Chatbot: whatever it happens to write. Grongy AI: 10 styles, regenerated from the saved metadata when you switch.

When a chatbot is the better tool

This is not a page arguing that you should never open ChatGPT. There are jobs it does better than we do, and pretending otherwise would be exactly the kind of marketing this product exists to avoid.

  • Thinking out loud — pressure-testing an argument before you commit to it.
  • Explaining an unfamiliar method, statistic or paper in plain language.
  • Quick, throwaway rewriting where no citation is involved.
  • Everything outside academic writing, which is most of what it is good at.

Questions before you start

Can I write my thesis with ChatGPT?

You can draft text with it, and many people do. The trouble appears at the two points a thesis is actually judged on: whether the references are real, and whether the claims are supported by them. Both need verification you have to do yourself — and the more text you generate, the more verification you owe.

Does ChatGPT invent references?

Language models can produce references that do not exist. It is well documented across published studies, at rates that vary widely by model and by topic. Treat every reference a chatbot gives you as unverified until you have opened the paper and matched the authors, year, title and venue.

Is Grongy AI just a wrapper around ChatGPT?

It uses frontier language models for the writing, as AI writing tools generally do. The difference is not the model — it is what the model is allowed to cite. Drafting reads the papers in your project library and cannot cite anything outside it, and a claim the library does not support stays uncited.

Do universities allow AI-assisted thesis writing?

Policies differ and are still changing. Most that permit it require disclosure and hold you responsible for the accuracy of everything you submit. Read your department's rule before you rely on any tool, including this one.

Explore more