3.5 Revision Techniques¶
What you will learn on this page
- A workflow where you draft, and AI acts as a diagnostic doctor, accumulating your common weaknesses
- A one-sentence improvement loop
- "Quiz" techniques to help you find errors yourself
- Paraphrasing (rewriting in different structures)
- A staged revision process (macro → meso → micro)
- Non-AI revision techniques (reverse outlining)
- Cross-section consistency checks
- How to revise with AI while keeping your voice
- How to avoid AI "homogenization"
- Final checks for the Abstract
- A final pre-submission checklist
- How to write a cover letter
- How to respond to reviewer comments
- AI-use disclosure and work logs
How this page fits in
Pages 3.1 through 3.4 covered section-by-section drafting and refinement.
This page focuses on whole-paper revision and submission preparation. For prompt principles, see 3.1.
Draft first, then diagnose: AI as a doctor¶
Write at least 200–300 words on your own first. Then give AI only a diagnostic task—not a rewriting task.
Identify up to 3 issues each in the following categories for the text below:
grammar/usage, logic, and clarity.
Provide only 1 suggested revision per issue. Keep explanations brief.
[Paste your draft here]
The key is to use AI for identifying weaknesses, not for "fixing everything." If you use the same diagnostic categories every time, your own weaknesses accumulate as data over time.
Logging weaknesses and reviewing patterns¶
Example: weakness log format
| Date | Section | Category | Issue | My understanding |
|---|---|---|---|---|
| 2/11 | Methods | Grammar | Tense inconsistency (present → past) | Procedures should use past tense consistently |
| 2/11 | Results | Usage | Unnatural placement of significant | Statistical significance uses adjective form |
| 2/11 | Discussion | Logic | Logical leap in causal claim | Data only shows correlation |
Prompt: analyze weakness patterns
Below is a list of issues that AI has pointed out in my English writing so far.
Identify recurring patterns (= my weaknesses) and for the top 3:
(1) Describe the pattern
(2) Explain why this mistake is common (e.g., L1 transfer from Japanese)
(3) Suggest what to watch for during revision
[Paste issue list here]
One-sentence improvement loop¶
Having AI revise an entire text at once diminishes the learning effect. Instead, submit just one sentence at a time.
Step 1: Generate candidates
Rewrite the following sentence to be more natural and academic,
without changing the meaning.
Provide 3 candidates. Briefly explain the difference for each.
[Paste target sentence here]
Step 2: Narrow down with the reader in mind
Of those 3 candidates, which one would be least likely to be
misread by an applied linguistics researcher? Explain in one line.
Step 3: Decide and adopt it yourself
The act of "choosing" is where learning happens. Do not let AI make the final decision—make the judgment yourself.
Turn revision into a quiz (to build your skill)¶
Rather than having AI proofread, this technique trains your ability to spot errors yourself.
Assume there are 5 unnatural points in the text below.
Give hints step by step.
Do not reveal the answers until I ask.
[Paste text here]
You guess first, then check your answers. It doesn't take long, but it reliably builds your skill.
Paraphrasing¶
The ability to express the same meaning in different ways is extremely important in academic writing.
Rewrite the following sentence in 3 different syntactic structures
without changing the meaning.
For each, indicate which structural pattern you used
(passive voice, participial construction, It-clause, etc.).
[Paste sentence here]
A staged revision process: macro → meso → micro¶
Rather than completing revision in one pass, it is more effective to change your focus across multiple rounds. The following three stages help you efficiently catch problems at different levels.
Stage 1: Macro level (structure and logic)¶
Check the overall structure and logical flow of the paper.
Check the manuscript below for macro-level issues.
Checklist:
(1) Does each section fulfill its role?
(2) Is there consistency across Introduction → Methods → Results → Discussion?
(3) Do the results and discussion correspond to the research questions?
(4) Is the paragraph order logical (are there paragraphs that should be rearranged)?
Do not correct grammar or wording. Point out structural issues only.
[Paste manuscript here]
Stage 2: Meso level (paragraph and sentence connections)¶
Check logical development within paragraphs and connections between paragraphs.
Check the manuscript below for paragraph-level issues.
Checklist:
(1) Is the topic sentence of each paragraph clear?
(2) Are sentence-to-sentence connections (Given-New, transition expressions) natural?
(3) Are transitions between paragraphs smooth?
(4) Is there unnecessary repetition?
[Paste manuscript here]
Stage 3: Micro level (grammar, usage, and expression)¶
Check individual sentences for grammar, usage, and expression.
Check the manuscript below for sentence-level issues.
Checklist:
(1) Grammar errors (subject-verb agreement, articles, prepositions, etc.)
(2) Appropriateness of tense
(3) Verbose expressions
(4) Unnatural collocations
(5) Spelling and punctuation
For each issue, quote the relevant passage and suggest one revision.
[Paste manuscript here]
Follow the stage order
It is important to revise in the order macro → meso → micro. If structural problems remain, grammar corrections at the sentence level may be wasted if the text is rewritten later.
Non-AI revision: reverse outlining¶
AI-powered revision is powerful, but combining it with non-AI techniques allows you to catch problems that AI misses.
Reverse outlining¶
This technique creates an outline in reverse by extracting the main claim of each paragraph from an already-written manuscript.
Procedure:
- Extract the topic sentence (main claim) of each paragraph in one sentence
- Read only the extracted sentences in order from top to bottom
- Check whether the logical flow is followable and there are no leaps
- Check whether any paragraphs repeat the same claim
This is fast and reliable
It reveals missing logical steps and redundant paragraphs quickly.
Read-aloud check¶
Reading your English text aloud makes it easier to catch problems that are hard to notice when reading silently:
- Sentences too long to read in one breath (consider splitting)
- Passages with unnatural rhythm (word order or structural issues)
- Unnecessary repetition of the same words or phrases
Cross-section consistency checks¶
Checking for consistency of terminology, numbers, and claims across the entire paper is extremely important.
Common inconsistency patterns¶
| Type of inconsistency | Example | Impact |
|---|---|---|
| Terminology variation | "EFL learners" in Methods, "English learners" in Results | Readers wonder if these refer to the same group |
| Participant number mismatch | N = 120 in Methods, N = 118 in Results (no explanation of exclusion) | Credibility of data is questioned |
| RQ–results misalignment | 3 RQs stated but only 2 answered in Results | Will certainly be flagged by reviewers |
| Abbreviation inconsistency | Full form missing at first use, different abbreviation used later | Causes reader confusion |
Prompt: cross-section consistency check
Check the following manuscript for consistency across sections.
Checklist:
(1) Terminology: Are different terms used for the same concept?
(2) Numbers: Are participant counts and variable values consistent across sections?
(3) RQ alignment: Are all RQs from the Introduction answered in Results/Discussion?
(4) Abbreviations: Are they defined in full at first use and used consistently thereafter?
(5) Figure/table numbering: Do in-text references match actual figure/table numbers?
If inconsistencies are found, quote the specific passages.
[Paste full manuscript here]
Revise with AI while keeping your voice¶
AI-generated text is fluent, but it tends to produce writing that "could have been written by anyone." To maintain your perspective as a researcher:
- Do not paste AI output directly—always write your own alternative first and then compare
- Place your version and the AI version side by side, and create the final version by mixing them
- Make a brief note of "why I chose this expression"
This alone reduces both the risk of wholesale copying and the loss of learning opportunities.
Avoiding AI "homogenization"¶
Generative AI tends to output "average academic English style." As a result, there is a risk of "homogenization" where every paper sounds the same.
Tips for avoiding homogenization:
- Maintain expressions characteristic of your field: Do not let AI over-generalize field-specific idiomatic expressions or jargon
- Always modify AI output—never use it as-is: Make at least one change per sentence based on your own judgment
- Compare outputs from multiple models: Compare ChatGPT, Claude, and Gemini outputs, and use the one closest to your own style as reference
Prompt: check stylistic consistency with your own writing
Compare the following two texts.
Text A is a section I wrote myself; Text B is a section I wrote with AI assistance.
Check for obvious stylistic differences (vocabulary level, sentence length,
frequency of hedging, passive voice usage, etc.).
If unnatural differences exist, suggest how to bring Text B closer
to the style of Text A.
Text A (self-written):
[Paste Text A here]
Text B (AI-assisted):
[Paste Text B here]
Prompt: revise but preserve voice
For more on Augmented Competence, see 1.2 Academic Integrity and Plagiarism Prevention.
Final checks for the Abstract¶
The Abstract is the "face" of the paper—many readers read it first (and sometimes only). As discussed in 2.1, it is typically written last. After completing revision of the main text, check the Abstract against the final version of the paper.
Prompt: Abstract and main text consistency check
Compare the following Abstract with the main text and check for consistency.
Checklist:
(1) Does the Abstract contain all 5 elements
(Background / Purpose / Methods / Results / Conclusion)?
(2) Does the Abstract match the main text
(especially numbers, key results, and conclusions)?
(3) Does the Abstract contain any information not in the main text (new claims)?
(4) Is the word count within the submission guidelines (target: [word count] words or fewer)?
Abstract:
[Paste Abstract here]
Main text summary (for each section):
- Introduction: [summary]
- Methods: [summary]
- Results: [summary]
- Discussion: [summary]
Final pre-submission checklist¶
A checklist of items to confirm before submitting your paper. For detailed checks specific to each section, refer to the relevant pages.
Content and structure¶
- Title accurately reflects the research content
- Abstract is within the required word count and contains all 5 elements (Background/Purpose/Methods/Results/Conclusion)
- Numbers and conclusions in the Abstract match the main text
- Keywords are appropriately selected
- All RQs in the Introduction are answered in Results and Discussion
- Discussion includes Limitations
- Conclusion does not introduce new arguments
Style and formatting¶
- Tense follows section-specific conventions (→ 2.3)
- Hedging expressions are appropriate (neither excessive nor insufficient)
- British/American spelling is consistent (-ize/-ise, -or/-our, etc.)
- All abbreviations are defined in full at first use
- Number formatting is consistent (spelling out at sentence beginnings, etc.)
- Terminology is consistent across sections
Figures, tables, and references¶
- Figure/table numbers match in-text references
- Figure/table captions are complete (→ 3.3)
- References conform to the target journal's style
- All in-text citations appear in the reference list (and vice versa)
- DOIs are included for references that have them
- Software citations are included in the References (→ 3.3)
Submission requirements¶
- Word count / page count within limits
- Font, margins, and line spacing as specified
- Author information anonymized (for blind review)
- Cover letter prepared
- AI-use disclosure statement prepared (→ 1.3)
- Data Availability Statement included (→ 3.3)
Prompt: comprehensive pre-submission check
Perform a final pre-submission check on the following manuscript.
Checklist:
(1) Abstract word count and coverage of 5 elements
(2) Correspondence between RQs and Results/Discussion
(3) Cross-section consistency of terminology and numbers
(4) Abbreviation definitions at first use
(5) Match between in-text citations and reference list
(6) Figure/table number correspondence
For each item, judge "OK" or "Needs review."
For "Needs review," point out the specific problem location.
[Paste full manuscript here]
Cover letter¶
Most journals require a cover letter when submitting a manuscript. The cover letter is an important document that conveys the paper's overview and your intent to the editor.
Elements to include in a cover letter¶
| Element | Content | Note |
|---|---|---|
| Addressee | Editor's name (if known) or Editorial Office | "Dear Editor," is acceptable |
| Manuscript title | Title of the submitted paper | Must match the manuscript exactly |
| Research overview | What was investigated, what was found (2–3 sentences) | A condensed version of the Abstract |
| Novelty and significance | Why this research is important | Value for the journal's readership |
| Fit with journal | Why you are submitting to this journal | Connection to the journal's scope |
| Declarations | Not under dual submission, all authors' consent, etc. | Per the journal's requirements |
| AI-use disclosure | Status of generative AI use (if required) | Follow the journal's policy |
Prompt: draft a cover letter
Draft a cover letter in English for submission to an academic journal,
based on the following information.
Requirements:
- Formal academic style
- 300 words or fewer
- Emphasize the novelty and significance of the research
- State the connection to the journal's scope
- Confident but not excessively self-promotional
Manuscript title: [title]
Target journal: [journal name]
Research overview: [2–3 sentences]
Key findings: [1–2 sentences]
Reason for submitting to this journal: [1–2 sentences]
Author(s): [author name(s)]
Differentiating your cover letter
A cover letter is not just a formality—it is an opportunity to make the editor think "I want to read this." Stating specifically "why this journal and not another" can make a positive impression.
Responding to reviewer comments¶
Responding to reviewer comments after peer review is an important part of the manuscript process. There are established patterns for writing response documents (Response to Reviewers / Rebuttal Letter).
Basic principles¶
- Respond to every comment individually: There is basically no comment you can ignore
- Polite but confident: No need to be overly humble, but show appreciation
- Specify where changes were made: State concretely what was changed and where in the manuscript
- Provide evidence when you disagree: Support your position with data or literature
Response templates¶
| Situation | Response pattern |
|---|---|
| Agreed with comment and revised | Thank you for this suggestion. We have revised the manuscript as follows: [description of change]. Please see page X, lines Y–Z. |
| Partially agreed | We appreciate this comment. While we agree that [point of agreement], we believe that [your view] because [evidence]. However, we have added [addition] to address this concern. |
| Disagree | We thank the reviewer for raising this point. We respectfully disagree because [evidence]. [Supporting data or literature]. We have added a clarification on page X to address potential concerns. |
Prompt: draft a response to reviewer comments
Draft a response in English to the following reviewer comment.
Requirements:
- Polite but confident tone
- Follow the order: appreciation → response → specify change location
- If a change was made, leave page/line numbers as [to be filled in]
- If no change was made, state the rationale clearly
Reviewer comment:
[Paste reviewer comment here]
Response plan (in brief):
[agree / partially agree / disagree + reason]
Prompt: organize reviewer comments
Read the following reviewer report and classify/organize the comments.
Categories:
(A) Content revision needed (research design, analysis, interpretation)
(B) Writing revision needed (style, grammar, formatting)
(C) Additional analysis or data needed
(D) Points where the author should push back (reviewer misunderstanding or difference of opinion)
For each comment:
- Category label
- One-sentence summary of the comment
- Difficulty of response (high / medium / low)
- Priority of response
[Paste reviewer report here]
Do not respond emotionally
Even after receiving harsh comments, always maintain a professional attitude in your response. Avoid expressions like "The reviewer is wrong" and use polite disagreement such as "We respectfully disagree." Having AI draft the response can help eliminate emotional language.
AI-use disclosure and work logs¶
The most common problem when using generative AI for writing is not the quality of the text itself, but being unable to explain afterward "which parts are my own work and which parts are AI-assisted." Keep responsibility anchored to the human side and ensure you can explain your process in a verifiable way.
Set a one-line personal rule first
Example: "I use generative AI only for structural advice, paraphrasing options, grammar explanations, alternative expressions, and checking. I do not use it for adding content, generating facts or citations, or fabricating sources."
Disclosure statement templates (English)¶
Include these near the Acknowledgements section of the paper or in the cover letter. Replace the bold portions to suit your situation.
A. Limited to improving clarity (conservative):
The author used a generative AI tool (ChatGPT, OpenAI) to improve clarity and readability of the manuscript, including proofreading and wording suggestions. All content, analyses, interpretations, and final decisions remain the author's responsibility.
B. Including structural advice and paraphrasing (practical):
The author used a generative AI tool (ChatGPT, OpenAI) for language editing and drafting support, such as outlining, paraphrasing options, and grammar checking. The author verified all statements and references and is fully responsible for the final manuscript.
C. Including natural-language rendering of methods descriptions:
The author used a generative AI tool (ChatGPT, OpenAI) to assist in converting technical procedures into natural-language descriptions and to refine wording. The author ensured accuracy of all methodological descriptions and verified all factual claims and citations.
D. Including analysis code generation support:
The author used a generative AI tool (Claude, Anthropic) for drafting support (language editing, paraphrasing, outlining) and for generating initial analysis code in R/Python. All generated code was reviewed, tested, and verified by the author.
For submission policies and disclosure examples, see:
1.3 Guidelines from Publishers and Educational Institutions