Written by: Marc-Oliver Gewaltig & Alessandra Giugliano

There has never been a better stocked toolbox for writers than there is today. With so much assistance on offer it is easy to become distracted from the most important point: AI can draft a paper, but it cannot write YOUR paper. You still need to translate your ideas, your research and your thinking into a clear story that readers will understand and remember. In this first instalment of our series on scientific writing in the AI age, we show you how to turn your research into a compelling narrative.

Great science depends on more than polished sentences. George Gopen and Judith Swan note that the purpose of scientific discourse is not the “mere presentation of information and thought” but rather its actual communication. Complexity of thought need not lead to impenetrable prose, clarity comes from understanding what your reader needs. 

Paul Halmos, reflecting on how to write mathematics, puts it bluntly: to communicate an idea well you must have something to say, someone to say it to and you must be willing to write, rewrite and re‑write it. Generative models cannot supply your idea. They cannot decide which finding is worth highlighting or how it fits within the larger conversation. If you outsource those decisions, the result will feel like cheap plastic, polished but empty.

This first instalment in a three‑part series shows you how to turn your research into a compelling narrative. We will draw on classic advice from “The Science of Scientific Writing” and “How to Write Mathematics” to help you structure and polish your manuscript. 

If you need to sharpen your research question before writing, our guide on developing a strong thesis statement is a great place to start. In later parts we will delve into the mechanics of each section of IMRaD and discuss how to use AI responsibly without losing your voice.

Messy Reality Versus the Linear Story

If you read a few journal articles you might believe that research proceeds neatly from an idea to an experiment to spectacular results. In reality most projects wander. Experiments fail, side questions arise and results lead you in unexpected directions. Your job as a scientific writer is to carve out a simple arc from question to answer. This is the hardest part of writing because it means deciding what not to include.

There are two ways to reconcile messy reality with the linear structure readers expect:

1. Write early and often. Start drafting as soon as you begin your project and revise as your research question evolves. This iterative practice forces you to articulate why you are doing each step and helps you spot gaps in your logic. Few scientists have the discipline to keep a living draft, but those who do find it easier to assemble a final manuscript.

2. If you wait until the end, you must excavate your most compelling story threads and ruthlessly edit out everything else. Re‑read your notes and results, look for the thread that connects your initial question to the outcome you care about and discard detours. 

You may discover that your project contains material for more than one paper; in that case, split it. Each paper needs its own story. AI tools can help here by generating quick summaries of different framings. Compare those drafts to see which narrative highlights your contribution best, but remember that only you can decide what the real story is.

If you are still at the stage of defining your topic or gathering sources, our step‑by‑step research guide explains how to go from an idea to a working bibliography and may help you formulate a focused question.

The IMRaD Framework & Importance of Your Title

Over the past two centuries, scientists have converged on a standard format for research articles: Introduction, Methods, Results and Discussion (IMRaD). Gopen and Swan point out that readers come to a paper with structural expectations. Confusion arises when experimental details creep into the results section or when interpretation and data intermingle. A well‑structured paper respects those expectations:

• Introduction: explains what you tried to find out, why nobody has done it before and why it matters

• Methods: provides enough detail for others to judge the validity of your work and its reproducibility

• Results: presents the fruits of your research through data using figures and tables, describing what you found without interpretation 

• Discussion: interprets what the results mean, shows how they answer your original question and compares them with previous work

Start with Your Title

Before you start drafting these sections, craft a title that summarises your key findings. A good title tells readers what you discovered, not just what you studied. 

For example, 

• Strong title: “Slime molds navigate without a nervous system using distributed decision‑making” signals both subject and discovery

• Weak title: “A study of slime mold behaviour in laboratory conditions” merely states a topic. 

Begin where your research ended rather than where it began. Write one or two sentences that capture your outcome. This becomes your working title and your north star: everything in your paper should support it. If a method, result or side experiment does not serve your main claim, set it aside for another manuscript.

For a detailed walkthrough of each section in this framework, along with examples and checklists, see our IMRaD format guide on our blog.

Create Your Scientific Paper Outline: Writing in Spirals

Most beginning writers tackle each section in order, but the parts of a paper depend on each other. Halmos suggests writing in spirals. In each loop, you touch every section, gradually expanding your outline. Here is a practical approach:

1. Write headings for each section. Create placeholders for the introduction, methods, results and discussion (and abstract, conclusion or appendices if your journal requires them).

2. Review the purpose of each section. 

   • Introduction: state your finding and why it matters

   • Methods: describe how you obtained your results

   • Results: present what you found

   • Discussion: interpret what it means

3. Outline each section with bullet points. Use complete sentences to capture key ideas, figures and data you plan to include. Do not worry about prose at this stage.

4. Check for coherence. Does the introduction set up the results? Does the discussion answer the questions posed in the introduction? Do the methods explain how you obtained the results? Look for gaps; they tell you which experiments, analyses or references you still need.

5. Iterate and revise. Flesh out your outline into paragraphs, add figures and tables, then revise. Each loop of the spiral adds detail and improves flow. Early writing makes this process smoother, and allows you to update your title and outline as research evolves.  If you are writing after completing your research, use the spiral to excavate and structure your story threads and build your title and outline around that.
   
   

Next, we will flesh out each section of your scientific paper. Treat them separately as you draft, but keep the story consistent across all sections.

Introduction and Discussion: Your Narrative Frame

The Introduction and Discussion are twin sections that frame your narrative. Write them early because they determine what belongs in the Methods and Results. 

Your introduction has three jobs:

1. Define the question. Explain what you were trying to find out. A clear research question or hypothesis signals to readers what to look for.

2. Identify the gap. Explain why nobody has answered this question so far. The gap may be technical (no existing method), conceptual (the connection has not been made) or contextual (conditions have changed).

3. Explain the significance. Make clear why answering this question matters—for your field, for a broader scientific problem or for society.

The discussion answers those points in reverse:

1. Answer the question. Show how your results address the question or hypothesis raised in the introduction.

2. Change the field. Explain how your findings alter the current understanding. Do they support, refine or contradict existing models? Where do they explain previous results, and where do they fall short?

3. Outline limits and next steps. Discuss where your conclusions are valid, where they are uncertain and what further work is needed.

Draft these sections first to establish your narrative frame. Do not fill them with technical details; readers care about what your findings mean more than how clever your methods were. Ensure that the question you raise in the introduction is answered in the discussion. If they do not align, you have a structural problem that must be resolved before moving on.

As Gopen and Swan emphasise, misplacing information forces readers to divert their energy from understanding content to untangling structure. Your job is to guide them smoothly from question to answer.

The union of these two sections demonstrates your understanding of the broader context. This is where you show you're not just a technician running experiments, but a scientist who understands how your work fits into the bigger picture.

Scientific progress often depends on interpretation. UNESCO’s Science and Culture report notes that scientists advance knowledge by challenging accepted explanations and proposing new ways of interpreting facts. Your discussion is where you demonstrate originality and insight. AI cannot provide this; it can only summarise existing knowledge. Use AI to test different ways of phrasing your argument, but decide for yourself what your results mean.

Methods and Results: Detail Without Distraction 

The Methods and Results sections document what you did and what you found. They are often easier to write because they follow from your research plan and notebook. 

Yet there are two common pitfalls: 

1. Including too much detail 

2. Mixing interpretation with data 

Use your title and narrative frame to decide what belongs here. If a procedure or dataset does not support your main argument, omit it or place it in an appendix.

In your Methods section, provide enough information for others to judge validity and replicate your work. Outline your experimental or analytical approach, materials, equipment and statistical techniques. Precise descriptions allow readers to follow your reasoning. If some details are very technical, move them to supplementary material. Could another researcher reproduce your work from your Methods section? If not, add more detail.

In your Results section, present your findings without interpretation. Use figures and tables to display data and accompany them with concise statements describing what the data show. Keep your narrative focussed on the main findings; do not bury them under secondary analyses. Separate presentation (results) from interpretation (discussion) to improve clarity. Some journals allow you to discuss each result immediately after presenting it; if you do this, signal transitions clearly so readers know when you are interpreting.

Putting it All Together: From Outline to Manuscript

Once your outline is populated with bullet points and data, expand each section into full sentences and paragraphs. It is useful to keep your outline document visible while writing so you can refer back to stay on your chosen story arc.

Write in spirals: flesh out content, then revise for coherence. During this process you will discover gaps, such as claims not yet supported by evidence, missing transitions or unanswered questions. Use these gaps to guide additional analyses or literature searches. Read, revise and repeat. A good paper often goes through many drafts.. 

Feedback makes your manuscript stronger. Share your draft with co‑authors and mentors. Ask colleagues outside your field to read it; if they understand your story, you have probably succeeded in making it clear. Language models can help with grammar and phrasing, but human readers provide conceptual feedback that AI cannot. Use AI tools to test alternative framings; ask a model to summarise your argument in different orders or voices—then decide which version communicates your idea best.

Build Your Writing Skills

Writing is a craft you learn by doing. Halmos reminds us that you must write, rewrite and re‑write.

Here are a few tips that help you develop your writing skills while helping you in your daily research activities:

1. Practice: set aside time each day or week to write about your research: what you did, why you did it and what you learned. 15-30 minutes are a good start. Keep a log of your writing sessions to build momentum.

2. Read widely: improving writing improves thinking. Study how papers outside your field make complex ideas accessible. How do they structure arguments? What makes their writing clear? Notice how they introduce a problem, develop an argument and draw conclusions. Compare clear articles with opaque ones and ask yourself why one is easier to follow.

3. Create multiple feedback loops: use tools for quick mechanical checks on grammar and spelling, but seek substantive feedback from mentors and peers. Thesify’s pre-submission review can give you feedback as often as you need and whenever you want. Test your writing on readers unfamiliar with your topic to ensure accessibility. 

4. Routine: don’t wait until inspiration hits you. Reserve time in your daily schedule for writing and meticulously stick to the schedule. Establish a routine and treat writing as part of your research process rather than something you do only at the end.

Remember: every iteration of practice makes you not just a better writer, but a better scientist and communicator.

Recommended Resources:

For deeper guidance, return to the classics. 

• Gopen & Swan: The Science of Scientific Writing–  shows how reader expectations shape comprehension

• P.R. Halmos: “How to write mathematics”-- argues that writing is about communicating an idea; you must have something to say and avoid trying to say too many things

These resources, though decades old, remain relevant and will help you refine your craft.

Why Writing Still Matters in the AI Age

In this first part of our series on writing scientific papers in the AI age, we have argued that your idea and interpretation are the irreplaceable core of your paper. AI tools can assist with editing, outlining and exploring alternative storylines, but they cannot supply your thesis. By respecting the IMRaD structure, starting with a clear title, outlining in spirals, framing your narrative through the introduction and discussion and being selective in your methods and results, you will produce a manuscript that communicates your contribution effectively.

Part 2 will look more closely at each section, including how to craft a compelling abstract, organise your introduction, report methods and results in detail and write a discussion that situates your findings in the broader literature. We will also show how to incorporate AI tools into your workflow safely. Part 3 will explore ethical considerations, disclosure policies and how thesify can support your research without compromising your voice.

For advice on how to use AI without losing your own style, see our ethical AI writing guide. It explains when and how to employ tools responsibly so that your ideas stay at the centre of your work.

Ready to put these tips into practice?

Test thesify on your own manuscript to see how AI feedback can help you refine structure, clarify your thesis and polish your draft without losing your voice.

Related Posts

• The Typical Structure of a Scientific Research Paper: This guide breaks down the widely used IMRaD scientific article format—Introduction, Methods, Results, and Discussion—and explains what each section is for, how to approach writing it, and why it matters for effective scientific communication.

• How Researchers Are Using AI to Gain New Perspectives on Academic Writing: Understanding how your research fits into the larger academic landscape is crucial for making a meaningful contribution. To truly understand where your ideas fit, you need a clear map of academic conversations. thesify’s SOTA feature offers researchers this essential perspective, helping you identify the connections, gaps, and opportunities that define your field.

Peer Review Demystified: Navigate the Academic Publishing Process: This article guides you through the peer review process, clearly outlining each step and providing practical strategies to successfully navigate it. You'll learn the types of peer review, how to prepare your manuscript effectively, respond constructively to reviewer feedback, and ethically use emerging AI tools to enhance your submission.