Gains:
- Ability to understand the IMRaD (Introduction-Method-Findings-Discussion) structure and write contextual and purposeful draft prompts for each section.
- Ability to use AI as a writing coach for argument flow, paragraph structure, and inter-chapter consistency
- Being able to understand that the text produced by artificial intelligence is not the researcher's own idea and data, and that the draft must be originalized with his/her own contribution.
Research only becomes science when it is written down. Academic writing is the art of transforming your findings into a standardized text that other researchers can understand and evaluate. This is the phase that many researchers struggle with the most and spend the most time: fear of the blank page, how the sections will flow, how to build the argument. AI can be a powerful writing coach and outline generator here. But the core rule of this unit is about originality: the text the AI produces is not your idea, data, and interpretation; It is a draft and must be customized with your own contribution. The AI text may be fluent, but it is the original scientific contribution, not fluency, that makes an article valuable.
IMRaD: skeleton of the article
The vast majority of empirical (data-driven) articles follow the IMRaD structure: Introduction, Methods, Results, and Discussion. Each section has a separate job, and specifying which section you are writing for when requesting a draft from the AI significantly improves the quality of the output.
The introduction is like a funnel: it starts with the broad context, narrows to the gap in the literature, and descends to your question and purpose. Its job is to persuade the reader to ask "why is this study necessary?" The method describes your study clearly enough for someone else to replicate it: participants, tools, procedure, analysis. Past tense and neutral language are used. Findings present what you found without interpretation: tables, numbers, tests. Discussion is the opposite of the funnel: interpreting the findings, relating them to the literature, pointing out limitations and future work, opening back to the broader meaning.
AI can produce outlines that fit the purpose of each chapter — but only if you give it the right context. You provide the space and your purpose for the introduction, your actual procedure for the method, your actual findings for discussion. You never let the AI “make up” findings; The findings come only from your data.
Caution: When you "print" the Methods and Results sections into the AI, you provide the actual procedure and actual numbers. Asking AI to "produce a reasonable finding" is data fabrication and one of the gravest ethical violations.
Argument and flow
An essay is not a series of paragraphs, but an argument: each section and each paragraph builds on the previous one and supports the conclusion. A good paragraph usually begins with a topic sentence, develops with evidence, and connects to the next idea. Consistency between sections (same terms, same abbreviations, non-contradictory expressions) increases the reliability of the text. AI is very useful in this structural work: checking the argument flow of a draft, marking broken passages, rearranging a messy paragraph, checking whether a section contradicts another.
An efficient method is to work from skeleton to text: first you write down the main points of each section (this is your original thought), then you ask the AI to turn this skeleton into flowing paragraphs, and finally you rewrite it in your own voice and in the correct terms. So the idea comes from you, the AI just helps with the format, and the text remains yours.
Abstract and title: showcase of the article
The most read part of an article is the abstract; most readers only see it. A good abstract repeats the paper in miniature: purpose, method, key findings (in numbers), and conclusion. It is usually 150-250 words and a single paragraph; In journals that require a structured summary, it is divided into headings. The summary should be written last, when the article is finished — because only then will you really know what you are saying. AI is good at drafting an abstract from a completed article; but because every number and claim in the summary must match the text exactly, you meticulously compare the output with your own findings. A finding not included in the summary or an exaggerated conclusion undermines trust in the most visible place.
Title and keywords also determine findability: a title that contains the right terms, neither too bold nor too bold, will ensure that your article reaches the right reader. AI may suggest several alternative sets of titles and keywords; you choose them based on your field's search habits and the actual scope of the article. An exaggerated (“revolutionary”) or misleading title receives the first criticism in the peer review process.
three mini cases
Case 1 — From blank page to draft. One researcher couldn't start the introduction for weeks. He gave the gap, purpose and 5 key resources he determined to the AI and asked for a draft in the funnel structure. The draft that arrived was rough but provided a start; the researcher rewrote it completely, adding his own emphases. The blockage was broken; Instead of waiting two weeks, a working entry appeared in one day.
Case 2 — Flow control. One PhD student's discussion section was disorganized; Those who read it were asking "where is it going?" He gave the text to the AI and said, "mark the breaks and weak transitions in the argument flow, do not add content." The AI showed three disconnected points. The student corrected the transitions with his own words. The idea of the text has not changed, but its flow has become stronger.
Case 3 — No fabrication. A student tried to tell AI to "complete the findings section" due to lack of time; The AI would suggest numbers that seem reasonable but are not real. The student kept noticing: writing down findings only from his own analysis output, using AI only to correct sentence flow. So it gained momentum, but not a single made-up number entered the text.
Copiable templates
1) Entry outline (funnel):
Your role: academic writing coach. Establish a funnel structure for an introduction draft: (1) broad context, (2) gap in the literature, (3) my purpose, (4) my research question. Use the information below; source or findingFABRICATION. Context: [...], Blank: [...], Purpose: [...], Question: [...]
2) From skeleton to paragraph:
Below are my main points for the discussion section, point by point. Transform them into flowing, connected paragraphs in an academic style. ADDING a new claim or source; just format my points.Articles: [paste your own articles]
3) Flow and consistency control:
In the following section, mark breaks in the argument flow, weak transitions, and term/abbreviation inconsistencies. ADD CONTENT, just list structural issues. Section: [paste text]
4) Topic sentence reinforcement:
For each of the following paragraphs, evaluate whether the topic sentence is clear and suggest an alternative topic sentence if it is weak. Do not change the content of the paragraphs. Text: [paste]
Weak prompt / Strong prompt
Weak prompt:
Write me a discussion section about my work.
There is no context and no findings; AI spits out generic, empty or made-up text.
Powerful prompt:
Your role: writing coach. Below are my ACTUAL findings and commentary notes, item by item. Turn these into a discussion section outline: first summarize the findings, then relate them to the literature (I will add the sources, leave a placeholder), then boundaries and future work. Fabricating new findings or sources. Notes: [items]
Section
job
What would you give to AI?
Login
Establishes space and purpose
Context, space, purpose
Method
repeatability
actual procedure
Findings
Presentation without comment
Real numbers/output
discussion
Comment and context
Actual finding + comment note
Common mistakes
- To fabricate a finding or source. Method/Findings are written only from your actual data; Fake data is the most serious violation.
- Leaving the AI sketch as is. Without your own contribution and voice, the text is not original.
- Printing without context. Without space, purpose and findings given, the output is hollow.
- Confusing departmental affairs. Interpreting the findings and leaking raw numbers into the discussion destroys the structure.
- Not checking for consistency. Inconsistency of terms and abbreviations undermines trust.
Tip: Write down your main points as 4-6 bullet points before printing each section into AI. These items are your original contribution; AI just streamlines them. This way, you gain speed and keep the text truly yours.
In summary
Academic writing follows a standard structure (IMRaD) and establishes an argument. AI is a powerful coach in overcoming the blank page, turning the skeleton into a paragraph, and controlling flow and consistency. But it is the original contribution that makes the article valuable; The AI manuscript should be originalized with your own opinion, data, and voice, and the finding is never made up. The shortest rule: the structure and flow are from the AI, the original idea and real data are from you.
Application task
Select a section of your paper (preferably the introduction or discussion). First, write your main points in 5-6 articles. Ask the AI to turn these items into flowing paragraphs (without adding new claims/sources). Rewrite the incoming draft in your own voice and change at least two entire sentences. Then find and fix the breaks with the flow control prompt.
checklist
- [ ] Did I write my main points for each chapter first?
- [ ] I gave the AI real context/finding, not allowed for fabrication?
- [ ] Have I customized the draft with my own voice and terms?
- [ ] Have I separated the department work (finding without comment, discussion with comment)?
- [ ] Have I checked the flow of argument and consistency of terms?