These 55 prompts cover the research workflow from first question to final talk: sharpening a question, planning a search, extracting data from papers you paste, structuring a literature review, designing and critiquing methods, writing each section of a paper, organising reviewer comments and drafting responses, preparing grants and ethics applications, and rehearsing hard questions. It does not include a prompt for writing peer reviews of other people's work, because some funders and publishers prohibit AI in review. A language model asked to "suggest key studies" can return references that do not exist. So the prompts that work on papers, abstracts, notes or references tell the model to use only the material you supply, never to invent an author, year, DOI, quote or statistic, and to mark gaps instead of filling them.
The pack is written for postgraduate researchers, early-career and established academics, and research staff. It has seven sections: question and search strategy, reading and literature review, methodology and analysis planning, academic writing, journals and peer review, grants and ethics, and talks, teaching and integrity. Our students pack covers coursework and study skills.
Each prompt sets a role, takes your inputs as {{VARIABLES}}, states its rules and fixes an output format, so it is written to work in ChatGPT, Claude and Gemini. Four include a worked example with fictional inputs. The AI drafts and checks; you decide, verify and take responsibility.
Research Questions & Search Strategy
These prompts help you shape a question and plan a search before you read anything. They test feasibility, build Boolean strings for the databases you have, write inclusion criteria two screeners could apply identically, and keep a search log someone else could repeat. None of them lists papers. They work with the terms and abstracts you supply.
Best for: Narrowing a broad topic into a question you can finish
You are an experienced doctoral supervisor. Help me sharpen a research question. Do not answer it.
Field: {{FIELD}}
My current question or topic: {{DRAFT_QUESTION}}
Constraints (time, data access, skills, word limit): {{CONSTRAINTS}}
Do this:
1. Say what the question currently claims, in one plain sentence, so I can see whether it matches my intent.
2. Test it against five criteria: answerable with evidence, specific enough to scope, feasible within my constraints, relevant to the field, and open (not yes/no or already settled). Give each a pass, weak or fail, with one sentence of reasoning.
3. List the vague terms that need defining before the question can be studied.
4. Offer three rewrites, each narrower or differently framed, and say what each one gains and gives up.
5. Ask me up to five questions whose answers would decide which rewrite to pursue.
Do not cite literature or claim what is already known about the topic; you cannot see my sources. If something depends on the state of the literature, say so and tell me to check it.
Output as numbered sections with a table for step 2.
Fill in: FIELD DRAFT_QUESTION CONSTRAINTS
Use before you commit months to a question, then again when your supervisor says it is too broad.
Best for: Turning a pile of abstracts into a testable gap
You are a research-methods reader helping me articulate a gap in the literature honestly.
Topic: {{TOPIC}}
Abstracts or notes I have read (each labelled with my own tag such as [P1], [P2]): {{PASTED_ABSTRACTS}}
Work only from the text above. Do not use outside knowledge about the field. Do not invent or add any reference, author, year, finding or statistic, and do not suggest that a gap exists in "the literature" at large, because you have seen only these items.
Tasks:
1. For each item, state in one line what was studied, in whom or what, and the main claim, tagged with my label.
2. List what these items collectively leave unaddressed, such as populations, contexts, methods, time frames or outcomes. Tag each point with the labels that support it.
3. Write two draft gap statements of two to three sentences each, worded as "In the sources I reviewed, ..." rather than as claims about the whole field.
4. List the searches I should run to test whether each gap is real, and what result would show it is not.
Output with headed sections, then the two drafts. End with a warning that a gap is only a hypothesis until I have searched properly.
Fill in: TOPIC PASTED_ABSTRACTS
Paste real abstracts you have read. Treat the gap as a hypothesis to test, not a finding.
Best for: Building and tuning database search strings
You are a research librarian with systematic-search experience. Build search strings I can test; do not run searches or list papers.
Research question: {{RESEARCH_QUESTION}}
Databases I can access: {{DATABASES}}
Field and any date or language limits: {{LIMITS}}
Steps:
1. Break the question into two to four concepts and list them.
2. For each concept, give synonyms, related terms, spelling variants (US and UK), acronyms, and broader and narrower terms. Mark which words are plain-language terms I am guessing at.
3. Write one Boolean string per database using AND, OR, quotation marks, truncation and proximity operators correctly for that database. Where you are unsure of a database's exact syntax or field codes, say so and tell me to check its help page.
4. Suggest a narrower and a broader version of each string, and what each would change.
5. List controlled-vocabulary headings (such as MeSH or thesaurus terms) only as ideas to look up. Do not state they exist in a given database unless I confirm.
Do not name any paper, author or journal. Output as a table per database followed by a short testing plan.
Fill in: RESEARCH_QUESTION DATABASES LIMITS
Test every string yourself and read the first fifty hits before you trust it.
Example output · Claude (example written by The Prompt Index)
Sample inputs (fictional)
- RESEARCH_QUESTION: Do school-based mindfulness programmes reduce test anxiety in secondary school students?
- DATABASES: PubMed, Scopus
- LIMITS: 2010 onwards; English; peer-reviewed
1. Concepts
A. Intervention: mindfulness programmes. B. Population: secondary school students. C. Outcome: test anxiety.
2. Terms (guesses at plain-language wording are marked)
| Concept | Terms |
|---|---|
| A | mindful*, meditat*, MBSR, MBCT, "mindfulness-based stress reduction", "breathing exercise*" (guess: broad, adds noise) |
| B | adolescen*, teen*, "secondary school*", "high school*", "middle school*", "year 9" (UK wording; guess) |
| C | "test anxiety", "exam anxiety", "exam stress", "examination stress", "evaluation anxiety" (US and UK variants) |
3. Strings by database
| Database | String |
|---|---|
| PubMed | (mindful*[tiab] OR meditat*[tiab] OR MBSR[tiab] OR MBCT[tiab]) AND (adolescen*[tiab] OR teen*[tiab] OR "secondary school*"[tiab] OR "high school*"[tiab]) AND ("test anxiety"[tiab] OR "exam anxiety"[tiab] OR "exam stress"[tiab] OR "examination stress"[tiab]) AND 2010:2026[dp] |
| Scopus | TITLE-ABS-KEY ( ( mindful* OR meditat* OR mbsr OR mbct ) AND ( adolescen* OR teen* OR "secondary school*" OR "high school*" ) AND ( "test anxiety" OR "exam anxiety" OR "exam stress" OR "examination stress" ) ) AND PUBYEAR > 2009 |
Limits: the Scopus string has no language limit, so apply the language refinement in Scopus's results sidebar, or look up the exact limit syntax in its search tips. PubMed's English filter is in its filters sidebar. Neither database has a reliable peer-reviewed filter, so peer-review status has to be checked per record or through the document-type options, which are not the same thing.
Syntax notes: in PubMed, operators are uppercase, a field tag goes after each term, and a wildcard needs at least four characters before the asterisk. In Scopus, only one wildcard is allowed per term. I am not certain of every platform feature, so check each database's help page.
4. Narrower and broader
| Version | Change | Effect |
|---|---|---|
| Narrower | Add randomi*[tiab] OR trial*[tiab] (PubMed); add randomi* OR trial* inside TITLE-ABS-KEY (Scopus) | Focuses on trials; may miss quasi-experimental studies |
| Broader | Replace concept C with (anxiety OR stress) AND (exam* OR test*) | Many unrelated results |
5. Vocabulary to look up
Check PubMed's MeSH database for headings on mindfulness and on test anxiety, and confirm their names before using [mh]. I have not stated that they exist.
Testing plan: run each string, read the first fifty titles, note what the irrelevant ones share, and log the date, platform and count for every run.
Best for: Writing criteria that two screeners would apply the same way
You are an evidence-synthesis methodologist. Help me write inclusion and exclusion criteria that two people could apply the same way.
Research question: {{RESEARCH_QUESTION}}
Type of review (systematic, scoping, narrative, other): {{REVIEW_TYPE}}
Practical limits (languages, dates, access, time): {{LIMITS}}
Tasks:
1. Suggest a fitting framework to structure the question (for example PICO, PEO, SPIDER or PCC) and fill it in, flagging every slot where my question is ambiguous.
2. Draft inclusion and exclusion criteria covering population, concept, context, study design, date, language and publication type. Give a one-line reason for each.
3. Rewrite any criterion that two screeners might read differently, and show a borderline example for it.
4. List the decisions I must record before screening starts, so I cannot adjust the criteria after seeing results without noting it.
5. Say which parts depend on guidance for my review type, and tell me to check the current method handbook or reporting standard rather than rely on you.
Do not cite any specific paper or handbook page numbers. Output as a numbered list with a final criteria table.
Fill in: RESEARCH_QUESTION REVIEW_TYPE LIMITS
Agree the criteria with a co-reviewer before you screen anything.
Best for: Keeping a reproducible record of what you searched
You are a careful research assistant. Help me keep a search log that makes my literature search reproducible.
Project: {{PROJECT_NAME}}
Searches I have run so far (database, date, exact string, filters, number of results; paste what you have, even if messy): {{SEARCH_NOTES}}
Reporting standard I must follow, if any: {{REPORTING_STANDARD}}
Tasks:
1. Turn my notes into a clean log table with columns: ID, database and platform, date run, exact string as typed, filters applied, results returned, notes.
2. Flag anything missing that a reader would need to repeat the search, such as the date, the platform, or limits applied. Write [MISSING] in the cell and list what to recover.
3. Spot strings that differ only slightly and say what changed, so I can describe the iteration honestly.
4. Draft a short methods paragraph describing the search, using only the facts in my notes. Mark every gap as [MISSING].
5. List what I should save at each search (exports, screenshots, dates) in case results change later.
Do not invent counts, dates, databases or settings. If my notes do not say it, it is not in the log.
Output as a table followed by the paragraph and a checklist.
Fill in: PROJECT_NAME SEARCH_NOTES REPORTING_STANDARD
Paste your messy notes straight after each search session, not at the end.
Best for: Finding the words other authors use for your topic
You are a research librarian helping me widen a search without losing focus.
My topic: {{TOPIC}}
Text from two to five relevant papers I already have, such as titles, abstracts, author keywords and subject headings (label each [S1], [S2] and so on): {{SEED_PAPER_TEXT}}
Using only the text above:
1. Extract every term the authors use for the same ideas: keywords, phrases, acronyms, and variant spellings. Group them by concept and tag each with the source label.
2. Point out where different sources use different words for the same thing, because these are the terms a single search would miss.
3. List terms that look like they come from other disciplines' vocabulary and might be worth searching.
4. Suggest which terms to try as precise search words and which are too general.
5. List the people, groups or journals named in my text as leads I might search by name later, without adding any I did not provide.
Do not add terms from memory and present them as coming from the sources. If you suggest an extra term of your own, put it in a separate list called "My suggestions, unverified".
Output as grouped lists, then a short table of best-bet terms.
Fill in: TOPIC SEED_PAPER_TEXT
Do this after your first few good papers, before the main search.
Best for: A plan that shows when it does not fit
You are a project-minded supervisor. Help me build a realistic plan from now to my deadline.
Project and the deliverable due: {{DELIVERABLE}}
Final deadline: {{DEADLINE}}
Today's date: {{TODAY}}
Work I need to do (stages, rough sizes, dependencies): {{STAGES}}
Fixed commitments and unavailable periods (teaching, leave, ethics board dates, fieldwork windows): {{COMMITMENTS}}
Produce:
1. A backwards plan with milestones and dates, calculating milestone dates from my deadline and today's date, and showing the working. Where a date depends on someone else (ethics approval, supervisor feedback, recruitment), mark it [EXTERNAL] and build in slack.
2. For each stage, key tasks, an editable estimate in weeks, and dependencies.
3. The critical path, in a sentence, and the three places the plan is most likely to slip.
4. A buffer of extra weeks for each risky stage, shown in its own column, with the reason.
5. If the plan does not fit before the deadline, say so plainly and offer three options for cutting scope or extending time, without pretending it fits.
Count in calendar weeks, say so, and subtract any holidays or leave I list under fixed commitments. Do not invent approvals, recruitment rates or processing times; ask me. Output as a table of milestones, then the risk list.
Fill in: DELIVERABLE DEADLINE TODAY STAGES COMMITMENTS
Revise it monthly. A plan that fails is easier to repair early.
Reading & Literature Review
Here the AI works only on papers you paste or have notes on. Extract data into a table, screen abstracts with reasons you can audit, triage a paper, outline an argument-led review, write a synthesis paragraph, explain a figure, explain why studies disagree, and appraise methods. The citation-accuracy prompt checks your sentence against the exact passage you cite, so over-claiming is caught before a reviewer catches it.
Best for: Pulling comparable data out of papers you hold
You are a meticulous research assistant extracting data from a paper for my review. You work only from the text I paste.
Paper text (methods, results and conclusion at minimum): {{PAPER_TEXT}}
What my review needs to know (for example sample, measures, effect, setting): {{EXTRACTION_FIELDS}}
Create a table with one row per field I listed plus these defaults: aim, design, sample and setting, measures, main findings, limitations the authors state, and funding or conflicts if mentioned.
Rules:
- Each cell must paraphrase or quote the text, with a short quote in quotation marks where numbers or definitions matter.
- Where the text does not say, write [NOT REPORTED]. Never fill a gap from general knowledge or by guessing.
- Do not add any reference, author name or year that is not in the pasted text.
- Keep the authors' claims separate from your observations. Put observations in a final section called "Things I would check", such as unclear denominators or results that seem to conflict with the abstract.
- Do not rate the paper's quality here.
Output as a table, followed by the "Things I would check" list.
Fill in: PAPER_TEXT EXTRACTION_FIELDS
Spot-check three cells against the PDF every time. If one is wrong, do not trust the rest.
Best for: First-pass screening with reasons you can audit
You are a second screener supporting my literature screening. You assist; I make every final decision.
Inclusion and exclusion criteria (numbered): {{CRITERIA}}
Records to screen, each with an ID, title and abstract: {{RECORDS}}
For each record:
1. Give a recommendation: INCLUDE, EXCLUDE or UNSURE.
2. Name the criterion numbers that drove the recommendation and quote the words in the abstract that show it.
3. If the abstract lacks information needed to apply a criterion, say exactly what is missing and mark UNSURE rather than guessing.
After the list:
- Tally the recommendations; I will recount them myself.
- List the criteria that caused the most UNSURE results, since those may need rewording.
- Remind me which records need full-text checks.
Do not use outside knowledge of the papers, do not invent details missing from an abstract, and do not add records of your own. Be conservative: when in doubt, UNSURE.
Process at most twenty records per reply and tell me where to resume. Output as a table with columns ID, recommendation, criteria, evidence quoted, and what is missing.
Fill in: CRITERIA RECORDS
Compare a sample against your own decisions and disclose AI-assisted screening in your methods.
Best for: Structuring a review around an argument
You are an academic writing coach. Help me outline a literature review that argues something, rather than lists papers.
Review question or purpose: {{REVIEW_PURPOSE}}
Word limit and type of document (thesis chapter, article section, standalone review): {{FORMAT}}
My notes, grouped by paper and tagged [P1], [P2] and so on: {{NOTES}}
Create an outline:
1. Propose a central line of argument for the review, in two sentences, based only on my notes.
2. Provide a section-by-section structure with a heading, the point each section makes, the tags of the papers that support it, and a suggested word allocation within my limit.
3. For each section, give one sentence on how it leads to the next.
4. Identify where I should compare, contrast or critique rather than summarise.
5. List what is missing from my notes to support each section, so I know what to read next.
Never add citations, authors, studies or facts of your own. If you think a point needs support I did not supply, write [NEEDS SOURCE] and describe the kind of evidence required, without naming any.
Output as a nested outline with a closing list of gaps to fill.
Fill in: REVIEW_PURPOSE FORMAT NOTES
Show the outline to your supervisor before you write full paragraphs.
Best for: Explaining a conflict between studies in your review
You are a critical appraisal tutor. Help me work out why studies I have read reach different conclusions.
Question both address: {{QUESTION}}
Study A notes (design, sample, measures, findings, as I recorded them): {{STUDY_A_NOTES}}
Study B notes: {{STUDY_B_NOTES}}
Using only my notes:
1. State each study's conclusion in one sentence.
2. Compare them on population, setting, definitions of key terms, measures, design, time frame, analysis and funding. Put the comparison in a table, writing [NOT IN NOTES] where I did not record something.
3. List the differences that could plausibly explain the disagreement, ranked from most to least likely, with the reasoning for each. Label these as possibilities, not conclusions.
4. Suggest what I should look up in the full texts to test each possibility.
5. Write a neutral paragraph I could adapt for a review, explaining the disagreement without taking a side.
Do not bring in findings, authors or references from outside my notes, and do not say which study is right.
Output as a table, a ranked list, then the paragraph.
Fill in: QUESTION STUDY_A_NOTES STUDY_B_NOTES
Use when two papers you plan to cite cannot both be right.
Best for: Critical appraisal notes you can defend
You are a sceptical methodologist reviewing a published study for my critical appraisal. Work only from the text I paste.
Methods and results text: {{PAPER_METHODS_RESULTS}}
Study type, if known (RCT, cohort, survey, qualitative, other): {{STUDY_TYPE}}
Appraisal checklist I must use, if any: {{CHECKLIST}}
Tasks:
1. Summarise the design in three sentences.
2. Assess, with quotes from the text, the main threats to validity for this kind of study: sampling and selection, measurement, confounding or alternative explanations, analysis choices, missing data, and generalisability. For qualitative work, use credibility, transferability and reflexivity instead.
3. Note where the conclusions go further than the reported methods and results support.
4. If I supplied a checklist, go through it item by item with yes, no or cannot tell, and the evidence in the text.
5. List what I should look for in supplementary material or the registration.
If the text does not give enough information for a judgement, say "cannot tell". Do not use outside knowledge about the authors or journal, and do not invent details of the study.
Output as headed sections. Close with a balanced paragraph on strengths.
Fill in: PAPER_METHODS_RESULTS STUDY_TYPE CHECKLIST
Appraisal is your judgement. Use the output as prompts for your own reading.
Best for: Catching over-claiming before a reviewer does
You are a citation-accuracy checker. I will give you sentences from my draft and the passages I based them on. Check whether each sentence is supported by the passage.
For each pair, I supply:
Claim sentence from my draft: {{MY_CLAIM}}
Passage from the source, copied from the paper: {{SOURCE_PASSAGE}}
Source label: {{SOURCE_LABEL}}
For each pair, return:
1. Verdict: SUPPORTED, PARTLY SUPPORTED, NOT SUPPORTED or CANNOT TELL FROM THIS PASSAGE.
2. The exact words in the passage that justify the verdict.
3. Any overstatement in my sentence, such as turning "associated with" into "causes", dropping a limitation, widening the population, or changing a number.
4. A more accurate rewrite that stays within the passage.
Rules: judge only from the passage. Do not use what you remember about the paper, and do not claim the paper exists or says anything beyond the passage. If the passage seems truncated, say so. Do not add references.
Output one block per pair, and keep the effect size and uncertainty in any rewrite where the passage gives them.
Fill in: MY_CLAIM SOURCE_PASSAGE SOURCE_LABEL
Run it on your most important claims, and always paste the real passage.
Example output · Claude (example written by The Prompt Index)
Sample inputs (fictional)
- MY_CLAIM: Peer tutoring improves reading outcomes for all primary-age pupils.
- SOURCE_PASSAGE: "In our sample of 214 pupils aged 7 to 9 across six schools, those who took part in twice-weekly peer tutoring for one term showed greater gains on the reading fluency measure than the comparison group (difference 4.1 words per minute, 95% CI 0.9 to 7.3). We found no clear difference on the comprehension measure. Because schools chose whether to take part, we cannot rule out differences between participating and non-participating schools."
- SOURCE_LABEL: [T3]
Source [T3]
1. Verdict: PARTLY SUPPORTED (the fluency finding for ages 7 to 9 is supported; the claim as written is much broader).
2. Words that justify it: "pupils aged 7 to 9", "reading fluency measure", "We found no clear difference on the comprehension measure", "we cannot rule out differences between participating and non-participating schools".
3. Overstatement:
- "all primary-age pupils" widens the population from ages 7 to 9 in six schools.
- "reading outcomes" is broader than fluency. The passage reports no clear comprehension difference.
- "improves" is causal. Schools chose to take part, which the authors say limits what can be concluded.
- The size and uncertainty of the effect are omitted.
4. More accurate rewrite: "In a study of 214 pupils aged 7 to 9 in six schools, those who took part in peer tutoring for one term gained 4.1 more words per minute on reading fluency than a comparison group (95% CI 0.9 to 7.3), with no clear difference in comprehension; because schools chose whether to take part, the authors cannot rule out differences between schools [T3]."
Note: this assessment rests only on the passage you pasted. I have not seen the full paper, so check whether the methods section describes how the comparison group was formed before you cite it.
Best for: Deciding which papers get your attention
You are a fast, careful reading assistant. Help me decide how much time a paper deserves. Work only from the text I paste.
My project and the question I need answered: {{MY_QUESTION}}
Paper text (abstract and introduction at minimum; add methods and conclusion if you have them): {{PAPER_TEXT}}
Give me:
1. A summary in three levels: one sentence, one short paragraph, and a bullet list of the main points. Use the paper's own claims and attribute them ("the authors report ...").
2. Relevance to my question: high, medium or low, with the sentences that justify the rating.
3. What the pasted text does not let me judge (design, sample, data, limitations), marked [NOT IN PASTED TEXT].
4. A recommended reading plan: read in full, read specific sections only, skim, or skip, with the reason.
5. Terms and background I need to understand it, listed as things to look up, not explained from memory.
6. Two questions to keep in mind while reading the full text.
Do not assess quality from the abstract alone, and do not add facts about the authors, the journal or other work. Do not add any reference. Output with the headings above.
Fill in: MY_QUESTION PAPER_TEXT
A summary is a map for reading, not a substitute for it.
Best for: Understanding a result outside your specialty
You are a patient colleague from a neighbouring discipline. Help me understand something in a paper I am reading.
My background and what I already know: {{MY_BACKGROUND}}
The passage, or a description of the figure or table, including axis labels, legends, captions and the numbers (typed or copied by me): {{PASSAGE_OR_FIGURE}}
What I am trying to work out: {{MY_CONFUSION}}
Do the following:
1. Restate the content in plain words, step by step, tied to the exact parts of what I pasted.
2. Define each technical term you use once, in a short phrase, and say whether the paper defines it that way or I should confirm it in the paper.
3. For a figure or table, explain what each axis, column and group shows, what pattern is visible, and what the authors appear to conclude. Say what the display does not show.
4. List the assumptions the author seems to make that a reader could question.
5. Give me three questions to test my understanding, with brief answer notes drawn from the text.
If the description of the figure is incomplete, say what is missing instead of guessing. If I upload an image, say which parts you could not see clearly. Do not add information from outside what I provided.
Fill in: MY_BACKGROUND PASSAGE_OR_FIGURE MY_CONFUSION
Check your understanding against the paper's own text, not only the explanation.
Best for: Turning notes into a paragraph that synthesises
You are an academic writing coach. Help me write one synthesis paragraph for a literature review, using only my notes.
The point this paragraph should make: {{PARAGRAPH_POINT}}
Notes on each source, tagged [S1], [S2] and so on, with what it found and its main limitation: {{SOURCE_NOTES}}
Write:
1. A paragraph of 150 to 200 words that opens with the point, then compares and connects the sources (where they agree, differ, or build on each other) rather than summarising them one by one, and ends by saying what the evidence collectively suggests and how firmly.
2. Cite sources only by the tags I gave, such as [S1]. I will replace the tags with real citations in my own style, so do not invent author names, years or page numbers.
3. Match the strength of each claim to my notes. Keep every hedge and limitation I recorded.
4. After the paragraph, list sentences that go beyond my notes, if any, and the evidence I would need to support them.
5. Offer one alternative way to order the sources.
If my notes do not support the stated point, say so and propose a point they do support. Output the paragraph, then the checks.
Fill in: PARAGRAPH_POINT SOURCE_NOTES
Replace the tags with real citations from your own reference manager only after rereading each source.
Methodology & Analysis Planning
Plan the study before the data exist. Compare designs and run a pre-mortem on yours, operationalise a construct, draft an interview guide, clean up a survey, build a first coding frame from excerpts, outline an analysis plan, and prepare a pre-registration, and get help with analysis scripts. Every prompt tells you which decisions belong to your supervisor, ethics committee or a statistician.
Best for: Choosing between designs with the trade-offs visible
You are a research-methods advisor. Lay out design options for my question; I will choose.
Research question: {{RESEARCH_QUESTION}}
Discipline: {{DISCIPLINE}}
Epistemic stance, if any (positivist, interpretivist, pragmatic): {{STANCE}}
Resources and constraints (time, budget, access to participants or data, skills): {{RESOURCES}}
Give me three feasible designs. For each:
1. Name the design and say in two sentences how it would answer the question.
2. The data it needs, where it might come from, and how much effort collecting it takes.
3. What claims it can support and what claims it cannot (for example causal versus descriptive).
4. Main threats to validity or trustworthiness, and one way to reduce each.
5. Ethical issues that would need attention.
6. A fit rating against my constraints: strong, moderate or weak, with a reason.
Then give a comparison table and tell me which two questions I should settle before choosing.
Do not cite methodological texts, papers or authors, and do not claim a design is the standard in my field. Say what to look up in my field's methods literature and ask my supervisor to confirm. If my question is unclear, ask for clarification before you design anything.
Output with headed designs followed by the table.
Fill in: RESEARCH_QUESTION DISCIPLINE STANCE RESOURCES
Take the table to your supervisor and ask which option the field would accept.
Best for: Finding weaknesses in your design while it is cheap to fix
You are a critical friend in a research design review. Assume my study has been run and the results are unconvincing. Work backwards to why.
Study summary (question, design, sample, measures, analysis): {{STUDY_SUMMARY}}
What a reviewer in my field would care about most: {{REVIEWER_CONCERNS}}
Tasks:
1. List the ten most likely reasons the study could fail or be rejected, grouped as design, sampling, measurement, analysis, practical and ethical risks.
2. For each, estimate likelihood and impact as low, medium or high, with one sentence of reasoning. These are judgement calls, not probabilities.
3. For the five most serious, propose a concrete change to the plan that reduces the risk, and say what it costs me in time or scope.
4. List the assumptions my plan makes that I have not tested.
5. Suggest a small pilot or check I could run in the next month to find out if the biggest risk is real.
Do not cite studies or invent evidence about how often such problems occur. If a point depends on conventions in my field, tell me to check with my supervisor or recent published methods sections.
Output as a risk table, then the five fixes, then the pilot.
Fill in: STUDY_SUMMARY REVIEWER_CONCERNS
Run this before ethics approval and again before data collection.
Best for: Defending your choice of measure in a methods section
You are a measurement specialist helping me turn an abstract concept into something measurable.
Construct: {{CONSTRUCT}}
Population and setting: {{POPULATION}}
Research question it serves: {{RESEARCH_QUESTION}}
Measures or definitions I already know of (names only, from my own reading): {{KNOWN_MEASURES}}
Do this:
1. Break the construct into dimensions or components and define each in plain language, noting where definitions in the field may compete.
2. For each dimension, describe the kinds of indicators or instruments that could measure it (self-report, behavioural, administrative data, observation, and so on) and the trade-offs of each.
3. Discuss validity and reliability concerns specific to my population and setting, and what evidence I would need to justify my choice.
4. Comment on the measures I listed: what each may capture and miss, based only on what I told you.
5. List the questions I must answer by reading the original validation papers.
Do not name or describe specific published scales unless I listed them, and never state their reliability figures, items or citations. If you think a well-known instrument may exist, say "search for established measures of X" instead. Output as a table by dimension, then the reading list questions.
Fill in: CONSTRUCT POPULATION RESEARCH_QUESTION KNOWN_MEASURES
Always read the validation paper for any scale before using it.
Best for: A draft guide ready for supervisor and ethics review
You are an experienced qualitative researcher. Draft a semi-structured interview guide.
Research question: {{RESEARCH_QUESTION}}
Participants (who they are, how they were recruited, likely sensitivities): {{PARTICIPANTS}}
Interview length and mode: {{LENGTH_AND_MODE}}
Theoretical or conceptual lens, if any (name it, or write none): {{LENS}}
Produce:
1. An opening script covering purpose, consent reminders, recording, confidentiality and the right to stop. Keep it conversational.
2. Four to six topic blocks in a sensible order, easy to difficult. For each block, give a main question, two to three probes, and the research question it serves.
3. Rewrite any question that leads, assumes, or asks two things at once, and show the original and the fix.
4. Flag questions that may cause distress, and suggest how the interviewer should respond.
5. A closing block that allows participants to add what I missed.
6. A short list of pilot-interview questions: where might the guide run over time or confuse people?
Do not claim the guide is validated. The guide must be piloted and approved by my ethics committee before use. Output as a formatted guide with timings.
Fill in: RESEARCH_QUESTION PARTICIPANTS LENGTH_AND_MODE LENS
Pilot with one or two people outside your sample before the first real interview.
Best for: Cleaning up questions before you send a survey
You are a survey methodologist reviewing my draft questionnaire. Do not change its purpose.
What I want to find out: {{SURVEY_AIM}}
Target respondents: {{RESPONDENTS}}
Draft questions, numbered, with answer options as I wrote them: {{DRAFT_QUESTIONS}}
For every question, check and report only the problems you find:
- leading or loaded wording,
- double-barrelled items (two questions in one),
- jargon the respondents may not know,
- answer options that overlap, leave gaps or lack a neutral or "prefer not to say" choice where appropriate,
- order effects from earlier questions,
- questions that do not link to the survey aim.
For each problem, quote the words, explain the issue in one sentence and give a rewrite.
Then:
1. Suggest an order of sections that reduces fatigue and sensitivity effects.
2. Estimate completion time by question count only, and note it is rough.
3. List questions I could cut.
4. Suggest three cognitive-interview checks to run with a few test respondents.
Do not invent standard scales or claim any wording is validated. Output as a table by question number, then the structure advice.
Fill in: SURVEY_AIM RESPONDENTS DRAFT_QUESTIONS
Test the revised version on a handful of people from your target group.
Best for: A first-draft codebook to test and revise
You are a qualitative analysis mentor. Help me draft a first coding frame from a small sample of data. The coding and interpretation remain mine.
Research question: {{RESEARCH_QUESTION}}
Approach (for example thematic analysis, content analysis, grounded theory): {{APPROACH}}
Anonymised excerpts, each labelled [E1], [E2] and so on: {{EXCERPTS}}
Tasks:
1. Propose an initial set of codes. For each code, give a short definition, an inclusion rule, and the labels of the excerpts it applies to, with a quoted phrase.
2. Mark whether each code is descriptive (what was said) or interpretive (what it might mean), and keep the two apart.
3. Point out excerpts that could take more than one code, or none, and explain the ambiguity.
4. Suggest how the codes might cluster into candidate themes, labelled as tentative.
5. List the questions I should ask to test the frame on a second sample, and how a second coder could check it.
Use only the excerpts. Do not invent quotations, participants or frequencies, and do not claim the codes are saturated. Remove any identifying details from my excerpts before pasting them, and confirm your ethics approval permits AI processing of the data. Output as a codebook table.
Fill in: RESEARCH_QUESTION APPROACH EXCERPTS
Use as a starting codebook only. Check data-handling permission before pasting any participant text.
Best for: Deciding the analysis before the data decide it for you
You are a statistical consultant helping me plan analysis before I collect data. You suggest options; you do not run tests on invented data.
Research question and hypotheses: {{HYPOTHESES}}
Design and variables (type and measurement level of each): {{VARIABLES}}
Sample size I expect and how I will recruit: {{SAMPLE_PLAN}}
Software I use: {{SOFTWARE}}
Produce an analysis plan outline:
1. For each hypothesis, name the plausible analysis approach, the assumptions it makes, and how I would check them.
2. Describe how I should handle missing data, outliers and multiple comparisons, as options with trade-offs.
3. Specify what should be decided before seeing the data, such as exclusion rules and the primary outcome.
4. Explain what effect size or interval I should report, in plain terms.
5. List the points where a statistician in my department should review the plan.
6. Explain what sample-size or power calculation I need, and what inputs it requires from me. Do not produce a number unless I supplied every input, and say the result must be checked.
Do not cite papers or claim a method is the accepted standard. Flag where choices are contested. Output as a table by hypothesis, then a checklist.
Fill in: HYPOTHESES VARIABLES SAMPLE_PLAN SOFTWARE
Take this to a statistician before you collect data, not after.
Best for: A clear pre-registration written before data collection
You are a research-transparency adviser. Help me draft a pre-registration from my design notes.
Registry or template I will use (for example OSF or another named template): {{REGISTRY_TEMPLATE}}
My design notes (question, hypotheses, sample, measures, procedure, analysis plans, exclusions): {{DESIGN_NOTES}}
Tasks:
1. Map my notes onto the usual pre-registration fields: question, hypotheses, design, sampling plan and stopping rule, variables, exclusion criteria, analysis plan, and what counts as confirmatory versus exploratory.
2. Draft each field in clear, specific language, using only what my notes state. Where my notes are vague or silent, write [DECISION NEEDED] and give the question I must answer.
3. Rewrite any hypothesis so it is directional or testable, and note where I changed meaning.
4. List the analytic choices that I could later change without noticing, and how to close them off in the text.
5. Give me a pre-submission checklist.
Never invent sample sizes, measures, effect sizes or decisions. Registry field names may differ from the template I named, so tell me to compare with the live form. Output as a field-by-field draft followed by the open decisions list.
Fill in: REGISTRY_TEMPLATE DESIGN_NOTES
Fill the [DECISION NEEDED] items yourself before you file anything.
Best for: Writing hypotheses that can be shown wrong
You are a methods tutor helping me turn a research question into hypotheses I could actually test. You do not predict results.
Research question: {{RESEARCH_QUESTION}}
Key variables and how I plan to measure them: {{VARIABLES}}
Design (experimental, correlational, qualitative propositions, other): {{DESIGN}}
Theory or reasoning I am drawing on, in my own words: {{MY_REASONING}}
Tasks:
1. State each hypothesis as a specific, falsifiable statement. For quantitative work give the null and alternative form, say whether the alternative is directional, and name the variables and the comparison. For qualitative work, write exploratory propositions or sensitising questions instead and say why hypotheses may not fit.
2. Link each hypothesis to the part of my reasoning that justifies it. If my reasoning does not justify a direction, say so and suggest a non-directional version.
3. For each, state what result would count against it, and what I would need to have measured to see that.
4. Flag hypotheses that are untestable with my design, that bundle two claims, or that restate the question.
5. Suggest which should be primary and which secondary, as options for me to discuss with my supervisor.
Do not cite theory or prior findings. If a hypothesis would need support from the literature, write [NEEDS SOURCE]. Output as a table with columns hypothesis, justification, disconfirming result, concerns.
Fill in: RESEARCH_QUESTION VARIABLES DESIGN MY_REASONING
Fix hypotheses before you collect data, then pre-register them if your field expects it.
Best for: Debugging and documenting analysis scripts safely
You are a patient data-analysis mentor for researchers who are not software engineers. Help with my analysis script without changing what the analysis is meant to do.
Language and packages (for example R with tidyverse, or Python with pandas): {{LANGUAGE_AND_PACKAGES}}
What the analysis should do, in plain words: {{ANALYSIS_GOAL}}
My code: {{MY_CODE}}
The error message or the unexpected output I see, copied exactly: {{ERROR_OR_OUTPUT}}
A description of my data columns and types, with no real participant data: {{DATA_DESCRIPTION}}
Do this:
1. Explain what my code does, line by line in plain language, and note any step that does not match my stated goal.
2. Explain the likely cause of the error or odd output, and suggest the smallest fix, showing the corrected lines only.
3. Point to silent problems: wrong joins, dropped rows, missing-value handling, factor or type issues, seeds not set.
4. Add comments that document each step, and suggest a way to check results (a count, a known case).
5. Tell me what I must verify myself, such as that the statistical test fits my design.
Do not invent output, numbers or results, and do not run claims about what my data show. Never ask for or accept identifiable participant data. If the cause is unclear, give two hypotheses and a test for each.
Fill in: LANGUAGE_AND_PACKAGES ANALYSIS_GOAL MY_CODE ERROR_OR_OUTPUT DATA_DESCRIPTION
Test any suggested code on a copy of your data and keep your original script under version control.
Academic Writing
Abstracts, introductions, methods, results, discussion and limitations, each built to keep your claims at the strength of your evidence. The editing prompts preserve numbers, hedges and citations, and flag what is missing rather than filling it in. A thesis argument map checks whether your chapters answer your research questions as one argument.
Best for: Writing an abstract that matches the paper
You are an academic editor who writes tight abstracts. Condense my manuscript into an abstract.
Journal or conference word limit and format (structured or unstructured): {{ABSTRACT_LIMIT}}
Manuscript text or detailed notes (aim, methods, key results with numbers, conclusion): {{MANUSCRIPT_TEXT}}
Audience: {{AUDIENCE}}
Write:
1. One abstract within the limit. If structured, use the standard headings for it (such as Background, Methods, Results, Conclusions).
2. Use only the findings, numbers and claims in my text. Copy every number exactly as it appears. Do not round, add or infer.
3. Make the conclusion no stronger than my text supports. Keep hedging where I hedged.
4. After the abstract, list any claim in it that I should double-check against the manuscript, and any important element of my text that did not fit.
5. Give a second version that is about twenty percent shorter and say what was cut.
Do not add citations or background statements from your own knowledge; include a citation only if the format requires one and my text supplies it. Output the abstract, the word count, the check list, then the shorter version.
Fill in: ABSTRACT_LIMIT MANUSCRIPT_TEXT AUDIENCE
Read the final abstract against the paper. It is the one part everyone will read.
Best for: Planning an argument-driven opening
You are a writing mentor for early-career researchers. Help me plan an introduction that moves from the broad context to my specific contribution. I will write the sentences.
Paper topic and aim: {{PAPER_AIM}}
Journal or field conventions I know about (for example IMRaD, length, tense): {{CONVENTIONS}}
Background points and sources I plan to use (each tagged [S1], [S2], and what each establishes): {{SOURCES_AND_POINTS}}
My contribution in one or two sentences: {{CONTRIBUTION}}
Tasks:
1. Propose a paragraph-by-paragraph outline: context, problem or tension, what is known (from my points only), the specific gap, my aim and approach, and a roadmap if the field uses one.
2. For each paragraph, give the main sentence, the evidence points to use by tag, and a transition to the next.
3. Warn me where a paragraph would read as a list of summaries, and how to turn it into an argument.
4. Flag claims that need a source I have not listed with [NEEDS SOURCE], without suggesting any particular paper.
5. Suggest a realistic length for each paragraph.
Never invent references, authors, dates or statistics. Output as a numbered outline plus a short note on tone for my field.
Fill in: PAPER_AIM CONVENTIONS SOURCES_AND_POINTS CONTRIBUTION
Use before drafting, then reuse the outline to check your finished introduction.
Best for: Cleaner methods that stay true to what you did
You are a copy editor with methods-section experience. Improve the clarity of my methods text without changing what was done.
Methods text: {{METHODS_TEXT}}
Reporting checklist I must follow (paste its headings; write none if not applicable): {{REPORTING_CHECKLIST}}
Tasks:
1. Rewrite the text for clarity, order and concision. Use past tense and a consistent order of events. Keep every number, instrument name, software version and procedural detail exactly as written.
2. Do not add any detail I did not write. Do not guess what was probably done.
3. Make a list of the things a reader would need to replicate or evaluate the study that my text does not state, marked [MISSING], such as exclusion criteria, ethics approval, consent, software settings, or how missing data were handled.
4. Compare my text with the checklist headings I pasted and say which are addressed, partly addressed or missing. If I wrote none, skip this step.
5. Flag sentences that are ambiguous about who did what.
Output the revised text first, then the [MISSING] list, then a short list of ambiguous sentences with questions for me. Show changes of meaning, if any, in a final "Please confirm" section.
Fill in: METHODS_TEXT REPORTING_CHECKLIST
Compare the revision against your lab notebook, not against memory.
Best for: Reporting results accurately and in order
You are a research writer who reports results precisely and without spin.
Research questions or hypotheses in order: {{HYPOTHESES}}
My tables or statistical output, pasted as text with their labels: {{RESULTS_TABLES}}
Reporting conventions for my field (for example how to present test statistics and intervals): {{REPORTING_CONVENTIONS}}
Write the results narrative:
1. One paragraph per research question or hypothesis. State the finding first, then the supporting numbers copied exactly from my tables.
2. Refer to tables or figures by their labels.
3. Report what was found, including null and unexpected results, in neutral language. Avoid terms like "highly significant" or "trend" unless my output supports them.
4. Keep interpretation out of this section unless the field combines results and discussion.
5. After the draft, list any numbers that look inconsistent or implausible in my tables (for example percentages that do not add up) so I can check them.
Do not compute new statistics unless I ask. Do not add any number, test or result that is not in my output. If something is missing, write [NOT IN OUTPUT]. Output the narrative, then the inconsistency list.
Fill in: HYPOTHESES RESULTS_TABLES REPORTING_CONVENTIONS
Recheck every number against the source output before it goes in the paper.
Best for: A discussion that does not overreach
You are a senior author advising on a discussion section. Help me structure it so the claims match the evidence.
Main findings (state each as in the results): {{MAIN_FINDINGS}}
Related work I want to discuss, with what each supports (tagged [S1], [S2]): {{RELATED_WORK_POINTS}}
Study limitations I am aware of: {{KNOWN_LIMITATIONS}}
Target journal's conventions, if any: {{JOURNAL_CONVENTIONS}}
Produce:
1. A skeleton: opening summary of the key finding, then each main finding interpreted, compared with the related work I listed, and explained, then implications, limitations, and conclusion.
2. For every finding, give the strongest claim my evidence supports and a stronger claim it does not support. Tell me to avoid the second.
3. Mark any place where I might be tempted to infer causation from correlation, generalise beyond the sample, or explain a null result as if it were a finding.
4. List alternative explanations a reviewer might raise for each key finding.
5. Mark with [NEEDS SOURCE] anything that would require evidence I have not supplied.
Do not add citations, authors or findings of your own. Output as an outline with notes under each heading.
Fill in: MAIN_FINDINGS RELATED_WORK_POINTS KNOWN_LIMITATIONS JOURNAL_CONVENTIONS
Use before drafting, then check each paragraph against its row.
Best for: Writing limitations that reviewers respect
You are an experienced journal reviewer who values frank limitations sections.
Study design: {{DESIGN}}
Sample: {{SAMPLE}}
Measures and analysis in brief: {{MEASURES_AND_ANALYSIS}}
Limitations I already know: {{KNOWN_LIMITATIONS}}
What I most want readers to still trust: {{CORE_CLAIM}}
Tasks:
1. Identify limitations a reviewer would likely raise that I have not listed, based only on the design details above. Rank them by importance.
2. For each limitation, state (a) what the limitation is, (b) which claims it affects, (c) which direction it might bias the result, if that can be reasoned from the design, and (d) anything I did to reduce it.
3. Write a limitations paragraph or two in measured academic tone. Be direct about the weaknesses. Avoid defensive language and excessive apology.
4. Suggest how future work could address each limitation, in one sentence each.
5. List claims elsewhere in the paper that the limitations may force me to soften.
Do not invent mitigations I did not carry out or cite methodological literature. If you are unsure whether a point applies, ask me. Output the ranked list first, then the draft text.
Fill in: DESIGN SAMPLE MEASURES_AND_ANALYSIS KNOWN_LIMITATIONS CORE_CLAIM
Share the ranked list with your co-authors, who may know of limits you missed.
Best for: Cutting length while keeping the argument
You are a copy editor for scholarly writing. Edit for clarity and concision only.
Field: {{FIELD}}
Audience: {{AUDIENCE}}
Word reduction target, if any: {{TARGET}}
Text to edit: {{TEXT}}
Rules:
- Preserve every claim, number, hedge, technical term, citation marker and quotation exactly. Do not change the argument, add content or strengthen or weaken any claim.
- Cut padding, repeated points, nominalisations, stacked prepositions and vague intensifiers.
- Prefer active voice where it reads naturally, but keep passive voice where it is conventional in my field.
- Keep my voice and keep the discipline's terminology.
- Do not insert references or "studies show" statements.
Return:
1. The edited text.
2. A table of the ten most significant changes, with before, after and the reason.
3. A list of any sentence where the meaning was ambiguous and you had to choose; ask me to confirm.
4. The word count before and after.
Show any sentence where tightening could alter the scientific meaning as "REVIEW" and leave it unchanged.
Fill in: FIELD AUDIENCE TARGET TEXT
Read the output next to the original. Reject any edit that shifts your meaning.
Best for: Consistent references without invented details
You are a careful copy editor checking reference formatting. You format what I give you; you do not look up, complete, correct or create references.
Citation style (exact name and edition, for example APA 7, Harvard, Vancouver, Chicago author-date): {{CITATION_STYLE}}
Style rules from my journal or department, if any: {{STYLE_RULES}}
My reference list, exactly as exported or typed: {{REFERENCE_LIST}}
In-text citations from my draft, if I want them cross-checked: {{IN_TEXT_CITATIONS}}
Tasks:
1. Reformat each entry into the named style using only the details I supplied. Never fill in a missing author, year, volume, pages or DOI. Write [MISSING: field] in its place.
2. Flag entries that look inconsistent or incomplete, for example a missing year or a journal name that appears in two forms. Do not say whether the work exists.
3. If I gave in-text citations, list any that have no matching entry and any entries never cited, matching by author and year as written.
4. List the style rules you applied and any you were unsure about, so I can check them against the official guide.
Remind me to check the formatted list against my reference manager and the original sources, because you can make formatting mistakes. Output the reformatted list, then the flags.
Fill in: CITATION_STYLE STYLE_RULES REFERENCE_LIST IN_TEXT_CITATIONS
Tidy formatting only. Confirm each entry against the source before submitting.
Best for: Checking that a thesis answers its own questions
You are an examiner-minded thesis adviser. Help me check whether my whole thesis hangs together as one argument.
Thesis title and central claim, as I would state it to an examiner: {{CENTRAL_CLAIM}}
Research questions in order: {{RESEARCH_QUESTIONS}}
For each chapter, its aim and a summary of what it concludes (label them [C1], [C2] and so on): {{CHAPTER_SUMMARIES}}
Discipline's expectations for a thesis, if I know them: {{THESIS_CONVENTIONS}}
Using only my summaries:
1. Draw the argument as a chain: what each chapter contributes, what it assumes from earlier chapters, and what the next chapter needs from it.
2. Check that each research question is answered in a named chapter and that the answer matches the central claim. List mismatches.
3. Find overlap between chapters, gaps between them, and conclusions that are stated before the evidence appears.
4. Identify the linking text each chapter transition needs (a closing paragraph and an opening paragraph), and give a one-sentence purpose for each.
5. List the three questions an examiner could ask about how the parts fit together.
Do not rewrite chapters, and do not comment on literature or evidence you have not been shown. If a summary is too thin to judge, write [SUMMARY TOO THIN]. Output as a chain table, then the mismatches, then the examiner questions.
Fill in: CENTRAL_CLAIM RESEARCH_QUESTIONS CHAPTER_SUMMARIES THESIS_CONVENTIONS
Run when most chapters are drafted, then again after your supervisor's feedback.
Journals, Peer Review & Responses
Write a cover letter that fits the journal's stated scope, turn a long decision letter into a work plan, draft courteous point-by-point responses tied to real changes, and test whether to push back. A mock review against the journal's criteria finds the first objections.
Best for: A clear, honest letter that fits the journal
You are an experienced corresponding author writing to a journal editor. Draft a cover letter for my submission.
Journal name and its stated aims and scope (copy from its website): {{JOURNAL_AIMS_AND_SCOPE}}
Manuscript title: {{MANUSCRIPT_TITLE}}
Article type: {{ARTICLE_TYPE}}
What the paper does, in three sentences: {{PAPER_SUMMARY}}
Why it suits this journal, in my own words: {{FIT_REASONS}}
Declarations to include (conflicts, funding, preprint, ethics, AI use): {{DECLARATIONS}}
Write a one-page letter:
1. A direct opening that names the manuscript and article type.
2. A short paragraph on the question, the approach and the main finding, matching my summary and not exceeding it.
3. A paragraph linking the paper to the journal's aims and scope as I stated them. Do not claim the journal has published related work unless I supplied the titles.
4. The declarations in a short closing paragraph, exactly as I have given them.
5. A courteous close.
Rules: no superlatives such as "groundbreaking" or "novel" unless my summary uses them and can be defended. Do not suggest reviewers, name editors, or mention impact factors or rankings. Do not invent facts. Output the letter, then three phrases I should check or soften.
Fill in: JOURNAL_AIMS_AND_SCOPE MANUSCRIPT_TITLE ARTICLE_TYPE PAPER_SUMMARY FIT_REASONS DECLARATIONS
Add editor names and reviewer suggestions yourself from the journal's submission page.
Best for: Turning a long review letter into a work plan
You are a calm, organised co-author helping me plan a revision.
Decision letter and reviewer comments, pasted in full: {{REVIEWER_COMMENTS}}
Manuscript word limit: {{WORD_LIMIT}}
Editor's revision deadline: {{REVISION_DEADLINE}}
Parts of the manuscript I can and cannot change (data fixed, analysis repeatable, and so on): {{WHAT_I_CAN_CHANGE}}
Tasks:
1. Number every distinct point in the decision letter and in each reviewer's comments, splitting compound comments into separate items. Keep the original wording in a column.
2. Classify each item: must fix, should address, clarification needed, disagreement, or out of scope.
3. State what each reviewer is really concerned about in one sentence, so I can see the intent behind the wording.
4. Identify comments from different reviewers that conflict or overlap.
5. Build a task list ordered by effort and dependency, with an estimate of the work in hours that I can adjust.
6. List items where I need co-author input or new analysis.
Do not draft responses yet, and do not decide whether a reviewer is right; give me the questions to decide that. Output as a table, then the conflicts, then the ordered plan.
Fill in: REVIEWER_COMMENTS WORD_LIMIT REVISION_DEADLINE WHAT_I_CAN_CHANGE
Do this before you reply to anything. A calm plan beats an angry first draft.
Best for: Courteous, specific replies that tie to actual changes
You are an experienced author who writes courteous, precise responses to peer review.
Reviewer comment: {{REVIEWER_COMMENT}}
What I changed in the manuscript, with page or line locations: {{CHANGES_MADE}}
Or, if I did not make a change, my reason and any evidence from my own data: {{REASON_NOT_CHANGED}}
Tone and journal conventions: {{TONE}}
Draft a response to this one comment:
1. Open by thanking the reviewer for the specific point, in one sentence, without flattery.
2. Restate the concern accurately in a neutral way.
3. Explain what I changed and where, using only the locations I gave you, and quote the new text if I supplied it. Or, where I did not change anything, give the reasoning politely and with evidence I supplied.
4. Do not claim any change, analysis or result that I did not describe. Do not add citations or numbers that I did not give.
5. Keep it concise and avoid defensiveness.
After the draft, give a short note: does my explanation answer the real concern, or only the wording? List what a reviewer could still push back on. Output the response, then the note.
Fill in: REVIEWER_COMMENT CHANGES_MADE REASON_NOT_CHANGED TONE
Do it one comment at a time. Keep the reviewer's text and your response side by side.
Example output · Claude (example written by The Prompt Index)
Sample inputs (fictional)
- REVIEWER_COMMENT: "The authors treat attrition as random, but 31 of the 120 participants dropped out before the follow-up. Please show that the results are not driven by who left."
- CHANGES_MADE: Added a comparison of baseline characteristics for completers and non-completers (new Table 3, p. 14, lines 301-312). Added a sensitivity analysis using last observation carried forward (p. 15, lines 322-330, Supplementary Table S2). Added a sentence to the limitations (p. 19, lines 410-413): "Attrition was 26 percent, and although completers and non-completers did not differ on the baseline measures we compared, dropout may relate to unmeasured factors."
- REASON_NOT_CHANGED: not applicable
- TONE: formal, collegial
[Reviewer #, Comment #]
Thank you for raising the issue of attrition.
We understand the concern to be that the 31 participants who left before follow-up may differ systematically from those who stayed, so that our estimates reflect who remained rather than the intervention.
We have made three changes. First, we added a comparison of baseline characteristics between completers and non-completers (new Table 3, p. 14, lines 301-312). Second, we ran a sensitivity analysis using last observation carried forward and report it in the text (p. 15, lines 322-330) and Supplementary Table S2. Third, we added a statement to the limitations (p. 19, lines 410-413): "Attrition was 26 percent, and although completers and non-completers did not differ on the baseline measures we compared, dropout may relate to unmeasured factors."
Note: Where could the reviewer still push back? Your changes show a comparison on observed baseline measures, which cannot address unmeasured differences, and last observation carried forward is a debated approach. A reviewer may ask for a method that models missingness, or ask what the sensitivity analysis showed. This reply says that the analysis was added but does not state its result, because you did not give me one. Add one sentence on what it found. The reply answers the concern about observed differences but only acknowledges the unmeasured ones.
Best for: Pushing back without burning bridges
You are a mentor who has handled many difficult reviews. Help me decide whether and how to push back on one comment.
Reviewer comment: {{REVIEWER_COMMENT}}
My view and the evidence from my own study or data that supports it: {{MY_POSITION_AND_EVIDENCE}}
What the editor has said about this point, if anything: {{EDITOR_GUIDANCE}}
Tasks:
1. Steelman the reviewer's point: state the strongest version of their concern.
2. List what the reviewer might be right about, even partly.
3. Judge whether my evidence addresses the concern, partly addresses it, or does not. Be candid.
4. If a partial concession is wise, suggest a change that satisfies the concern without compromising the work.
5. Draft a rebuttal of up to 150 words that is polite, concrete and based on the evidence I gave you. Offer a clarification in the manuscript when this is appropriate.
6. List what I should not say.
Do not invent evidence or cite any literature in support. If a stronger case needs outside support, write [I NEED A SOURCE] and describe the kind of evidence required. Output the analysis in order, then the draft.
Fill in: REVIEWER_COMMENT MY_POSITION_AND_EVIDENCE EDITOR_GUIDANCE
Run when a comment strikes you as wrong. The steelman step often changes your mind.
Best for: Finding what a reviewer would find first
You are a peer reviewer for a journal in my field. Review my manuscript as a critical but fair referee before I submit.
Journal aims, scope and reviewer criteria (copied from its website): {{JOURNAL_CRITERIA}}
Manuscript text: {{MANUSCRIPT_TEXT}}
Write the review in the format a journal would expect:
1. A neutral summary of what the paper claims to do.
2. Major concerns, numbered, each with the location in the text, why it matters and a suggested remedy.
3. Minor concerns.
4. How well the paper fits the criteria I pasted, scored criterion by criterion with a sentence of evidence from the text.
5. Claims that go beyond the evidence shown.
6. What the paper does well.
7. A tentative recommendation (accept, minor revision, major revision, reject), labelled as an impression.
Use only my text and the criteria. Do not claim the paper duplicates or contradicts literature you have not been shown, and do not invent references or methods standards. If you think the work should be compared with prior studies, say what kind of studies to check. Output as a structured review.
Fill in: JOURNAL_CRITERIA MANUSCRIPT_TEXT
Use it on your own work only. Never paste someone else's manuscript.
Grants, Ethics & Data Management
Aims and logic chains, lay summaries, a scoring pass against the funder's own criteria, ethics risk tables, participant information sheets, recruitment and consent text, data management plans and a resubmission response to panel feedback. These prompts work from the call text and approved protocols you paste, and mark every rule or regulation you must check yourself.
Best for: Aims a panel can follow in the first page
You are an experienced grant mentor. Help me turn a research idea into clear aims, objectives and a logic chain.
Funder and scheme: {{FUNDER_SCHEME}}
Call priorities, pasted from the call: {{CALL_PRIORITIES}}
Project idea in a paragraph: {{PROJECT_IDEA}}
Duration and team: {{DURATION_AND_TEAM}}
Produce:
1. A single overarching aim in one sentence.
2. Three to four specific objectives, each measurable or checkable, each linked to a work package and a likely output.
3. A logic chain that shows how activities lead to outputs, outcomes and wider impact, marking where each link is an assumption.
4. A table of risks to each objective with a mitigation.
5. A check against the funder priorities I pasted: where my project clearly fits and where the fit is a stretch.
6. Questions a panel would likely ask.
Do not invent preliminary data, track record, partners or statistics, and do not claim anything about the funder's rules beyond what I pasted. Where the scheme's rules matter (eligibility, page limits, costs), say that I must read the current guidance. Output as headed sections.
Fill in: FUNDER_SCHEME CALL_PRIORITIES PROJECT_IDEA DURATION_AND_TEAM
Share the aims with a colleague outside your subfield to see if they make sense.
Best for: A summary non-specialists can actually understand
You are a science communicator. Write a plain-language summary of my project.
Audience (funder panel lay member, public, patients, participants, policymakers): {{AUDIENCE}}
Word limit: {{WORD_LIMIT}}
Technical summary or abstract: {{TECHNICAL_SUMMARY}}
Write:
1. A summary that explains what the work asks, why it matters, what we will do, and what difference it could make, in plain words at a reading level for the audience.
2. Replace jargon or define it briefly. Use short sentences and concrete examples drawn from my text.
3. Keep every claim at the strength I wrote. Do not promise cures, breakthroughs, cost savings or timelines that my text does not state. Use "could" or "aims to" where I was uncertain.
4. Do not add facts, numbers or background that are not in my summary.
5. Give a one-sentence version for use in a tweet or a press line.
After the summary, list any technical terms you simplified and what nuance may have been lost, so I can check. Output the summary, the one-sentence version, then the list.
Fill in: AUDIENCE WORD_LIMIT TECHNICAL_SUMMARY
Ask someone outside your field to read it and say what they think the project is.
Best for: Anticipating what an ethics committee will ask
You are a member of a research ethics committee advising an applicant before submission. I will take your advice to my own committee, whose requirements prevail.
Study summary (aim, design, participants, procedures, data): {{STUDY_SUMMARY}}
Participant groups, including any who are vulnerable: {{PARTICIPANT_GROUPS}}
Institution and country: {{INSTITUTION}}
Ethics framework or committee guidance I must follow (paste or name it): {{ETHICS_FRAMEWORK}}
Produce:
1. A table of ethical risks: harms to participants, burden, consent and capacity, confidentiality and data protection, deception if any, power relations, researcher safety, and risks to third parties.
2. For each risk, give a plain description, who is affected, how likely and serious it may be (judgement only), and a proportionate mitigation I could put in the application.
3. Questions the committee is likely to ask about consent, withdrawal, data storage and sharing, and incidental findings.
4. Items that depend on law or regulation (data protection, medical research regulations) marked [CHECK RULES] with the kind of source to consult. Do not state specific legal requirements as fact.
5. Items I have not told you that you need to see.
Do not claim approval is likely. Output the table, then the question list.
Fill in: STUDY_SUMMARY PARTICIPANT_GROUPS INSTITUTION ETHICS_FRAMEWORK
This is preparation for the form, not a substitute for your committee's guidance.
Best for: Information sheets people can understand
You are a research ethics officer who writes clear participant materials.
Study purpose and procedures, from my protocol: {{STUDY_DETAILS}}
Participants (age group, reading level, language): {{PARTICIPANTS}}
Time commitment and payment or incentive, as in my protocol: {{TIME_AND_INCENTIVE}}
Risks and benefits as stated in my protocol: {{RISKS_BENEFITS}}
Data handling as approved by my institution: {{DATA_HANDLING}}
Contact details to print: {{CONTACTS}}
Write a participant information sheet that:
1. Uses short sentences and headings phrased as questions (What is this study? Why am I asked? What will I do? Can I stop?).
2. States that taking part is voluntary and that participants can withdraw, and explains what happens to their data if they do, using only what my protocol says.
3. Describes risks and benefits honestly, without overstating benefits.
4. Explains confidentiality and data storage using my approved details.
5. Gives contact details exactly as I supplied them.
Do not add promises, guarantees or details not in my protocol. Where the protocol is silent on something a participant would reasonably ask, write [PROTOCOL NEEDED]. Finish with a reading-level estimate and a list of terms to test with a lay reader. The text needs ethics committee approval before use.
Fill in: STUDY_DETAILS PARTICIPANTS TIME_AND_INCENTIVE RISKS_BENEFITS DATA_HANDLING CONTACTS
Match it to your approved protocol line by line before submitting for approval.
Best for: A first draft that identifies the decisions left to make
You are a research data steward. Draft a data management plan from my description. I will verify it against my institution's policies.
Project description: {{PROJECT}}
Data types and volume: {{DATA_TYPES}}
Funder or institution template headings, if I have them: {{TEMPLATE_HEADINGS}}
Sensitivity of data (personal, commercial, none) and storage I can use: {{SENSITIVITY_AND_STORAGE}}
Draft each section:
1. Data description: types, formats, volume (use my estimates only).
2. Documentation and metadata: what I should record so the data can be understood later.
3. Storage, backup and security, using only the systems I named.
4. Ethics, consent and legal issues for sharing.
5. Sharing, preservation and licensing: options for repositories and licences described by type. Do not name a specific repository as approved unless I did.
6. Roles and responsibilities, with placeholders for names.
7. Costs to include in the budget, as categories without amounts.
Mark anything that depends on law, funder policy or institutional rules as [CHECK POLICY]. Write [DECISION NEEDED] where I have not given enough to proceed. Do not invent retention periods, repository terms or regulations. Output using the template headings if provided, otherwise the headings above.
Fill in: PROJECT DATA_TYPES TEMPLATE_HEADINGS SENSITIVITY_AND_STORAGE
Ask your research data office to review it. They usually have a template.
Best for: Responding to panel feedback with changes, not defence
You are a senior grants mentor helping me plan a resubmission after a rejection or a request for revision.
Funder's feedback or panel comments, pasted in full: {{PANEL_FEEDBACK}}
My original aims and approach in brief: {{ORIGINAL_PROPOSAL}}
What has changed since (new data, publications, partners, pilot work), only things that are true: {{WHAT_CHANGED}}
Resubmission rules from the funder, pasted: {{RESUBMISSION_RULES}}
Tasks:
1. Split the feedback into separate points and group them as fatal flaws, fixable weaknesses, and comments of taste. Keep the funder's words in one column.
2. For each point, state what the panel likely wanted to see and propose a change to the proposal that would address it, using only what I told you has changed.
3. Identify points I should rebut politely instead of accepting, with the evidence I must supply.
4. Draft an introduction to the resubmission responding to the feedback, in the format the rules allow, with [DETAIL NEEDED] wherever the facts are missing.
5. Tell me honestly whether the feedback suggests a fundamental rethink rather than revision.
Do not invent pilot results, partners, publications or funder policies. Tell me to check the resubmission rules and deadlines with the funder. Output as a table by feedback point, then the draft.
Fill in: PANEL_FEEDBACK ORIGINAL_PROPOSAL WHAT_CHANGED RESUBMISSION_RULES
Wait a few days before you start, then read the feedback with the plan beside you.
Best for: Recruitment and consent text that matches the approved protocol
You are a research ethics officer helping me draft participant-facing recruitment and consent text. My ethics committee must approve everything before use.
Study summary from my approved protocol: {{PROTOCOL_SUMMARY}}
Who may take part (inclusion and exclusion criteria): {{ELIGIBILITY}}
How and where I will recruit (channels, platforms): {{RECRUITMENT_CHANNELS}}
Data handling and withdrawal terms exactly as approved: {{APPROVED_TERMS}}
Contact details to print: {{CONTACTS}}
Produce:
1. A short recruitment advert, within the character limit of a social post or poster, stating who I am, what taking part involves, the time required, eligibility and how to respond. No pressure language, no promises of benefit that my protocol does not state.
2. A consent form with numbered statements the participant ticks or initials (voluntary participation, right to withdraw and what happens to data, recording, anonymity limits, data storage, use of quotations or findings, contact for questions), phrased in plain language and using only the approved terms.
3. Notes on anything in the draft that my committee may query, such as incentives, vulnerable groups or online recruitment.
4. A list of items in the draft that must match the information sheet.
Do not invent terms, legal bases or storage periods. Write [PROTOCOL NEEDED] where the approved text is silent.
Fill in: PROTOCOL_SUMMARY ELIGIBILITY RECRUITMENT_CHANNELS APPROVED_TERMS CONTACTS
Compare every statement with your approved protocol before using it with participants.
Talks, Teaching & Research Integrity
Conference abstracts, a ten-minute talk plan with timings, a poster content plan, a hostile panel rehearsal for talks and vivas, a seminar plan for teaching, a tight update for your supervisor, a research statement for job applications, and an AI-use statement mapped to a publisher's policy.
Best for: An abstract that fits the call and stays honest about progress
You are an experienced conference programme committee member who also presents regularly.
Call for papers text (theme, format, word limit, selection criteria): {{CALL_FOR_PAPERS}}
My work (aim, method, findings so far, what is still to come): {{MY_WORK}}
Stage of the work (planned, in progress, complete) and what will be ready by the conference date: {{STAGE}}
Tasks:
1. Write an abstract within the stated limit that fits the call's theme and format. Describe the work at the stage it has actually reached.
2. If results are not final, say what will be presented rather than implying findings. Never write results that are not in my notes.
3. Use only the facts in my description. Do not add numbers, sample sizes or claims.
4. Show how the abstract meets each selection criterion in the call, in a short table with the sentence that does it.
5. Suggest a title of up to twelve words, and three alternatives.
6. List the questions a committee might ask about feasibility, novelty and fit, and which are weak spots.
Do not say the work is "novel" or "significant" unless my notes justify the word. Output the abstract, the criteria table, the titles and the weak spots.
Fill in: CALL_FOR_PAPERS MY_WORK STAGE
Check dates, word counts and the submission portal's rules yourself.
Best for: A talk with one clear message and honest timing
You are a presentation coach for researchers. Build a talk plan; I will make the slides.
Conference and session type: {{CONFERENCE}}
Audience: {{AUDIENCE}}
Talk length in minutes, excluding questions: {{TALK_MINUTES}}
Question time in minutes: {{QUESTION_MINUTES}}
Paper or project summary (question, method, key findings, limits, implications): {{PROJECT_SUMMARY}}
The one thing I want people to remember: {{TAKEAWAY}}
Produce:
1. A slide-by-slide outline with a title that states the point, the content, a suggested visual, and the speaking time. The speaking times must add up to my talk length exactly; show the sum.
2. A spoken opening of two or three sentences and a closing that returns to the takeaway.
3. Which results to show and which to leave out, with reasons, based on my summary.
4. Jargon to define for this audience.
5. Notes on pace: where to slow down and where to cut if I overrun.
6. Three backup slides to prepare in case of questions.
Use only the content I gave you. Do not add statistics, references or claims. If a figure is needed, tell me what it should show rather than describing results I did not state. Output as a table with columns slide, point, visual, time.
Fill in: CONFERENCE AUDIENCE TALK_MINUTES QUESTION_MINUTES PROJECT_SUMMARY TAKEAWAY
Time yourself aloud once. The plan is a guess until you have.
Best for: A poster that works for a viewer walking past
You are a poster design coach. Create a content plan for a research poster; I will do the layout.
Event: {{EVENT}}
Poster size and orientation: {{POSTER_SIZE}}
Project summary (question, methods, results, conclusions): {{PROJECT_SUMMARY}}
Likely viewers and how long they will stop: {{VIEWERS}}
Provide:
1. A headline of up to fifteen words that states the finding or question, with two alternatives.
2. The poster's blocks in reading order (introduction, question, methods, results, conclusion, next steps, contact), each with a word budget summing to a total I can print at readable size.
3. The text for each block, drawn only from my summary, in short phrases or sentences.
4. Which figure or table I should make the visual centre, what it should show and how to label it.
5. What to cut, and what to keep for conversation.
6. A thirty-second spoken pitch and a two-minute version.
7. A checklist for legibility, colour contrast and accessibility.
Do not add results, numbers or references that are not in my summary. If a block lacks content, write [MY INPUT NEEDED]. Output block by block.
Fill in: EVENT POSTER_SIZE PROJECT_SUMMARY VIEWERS
Print it on A4 first and read it from two metres away.
Best for: Practising the questions you dread
You are a panel of three senior academics in {{FIELD}}: a methodologist, a domain specialist and a sceptic from a neighbouring discipline. Run a rehearsal of questions for my talk or viva.
My abstract or summary: {{SUMMARY}}
Setting (conference talk, thesis viva, job talk, grant interview): {{SETTING}}
My weakest points, as I see them: {{WEAK_POINTS}}
Process:
1. Ask me one question at a time, alternating panel members, starting gently and escalating. Speak as the panel member, then stop and wait for my answer.
2. After my answer, step out of role with a line beginning "Coach:" giving two lines of feedback: what worked, and what a real panellist would still press on. Do not praise by default. Then ask the next question as the panel.
3. Include at least three questions about my weakest points, two about my methods, two about contribution and one about what happens if I am wrong.
4. After eight questions, summarise patterns in my answers: where I hedge, ramble or dodge.
5. List the questions I answered poorly, with notes on how to prepare.
Base all questions on my summary. Do not claim that known studies contradict me; if a question needs outside knowledge, phrase it generically. Start with the first question.
Fill in: FIELD SUMMARY SETTING WEAK_POINTS
Answer aloud, then type a summary. Stop and repeat on questions you fumbled.
Best for: A session plan that gets students talking
You are a university teaching developer helping a researcher plan a session.
Course and level: {{COURSE}}
Learning outcomes for the course: {{COURSE_OUTCOMES}}
Session topic: {{SESSION_TOPIC}}
Session length: {{SESSION_LENGTH}}
Class size and room or online format: {{FORMAT}}
Required readings, which I will name myself: {{READINGS}}
Produce:
1. Two to three learning goals for the session in verbs students can demonstrate, aligned to the course outcomes.
2. A timed plan with activities that need students to think and talk, not just listen: a starter, a core activity, a check of understanding, and a close.
3. For the core activity, give discussion questions, a task, and what a good response would show.
4. Common misunderstandings students may have, framed as things to listen for.
5. A short plan for quieter students and for dominant ones.
6. Preparation tasks for students beforehand, within a realistic workload.
Use only the readings I named. Do not describe their content or arguments from memory; leave placeholders such as [KEY POINT OF READING 1] for me to fill. Do not invent references. Output as a timeline table with notes.
Fill in: COURSE COURSE_OUTCOMES SESSION_TOPIC SESSION_LENGTH FORMAT READINGS
Fill the reading placeholders yourself from the actual texts.
Best for: Meetings that spend time on decisions
You are my research chief of staff. Turn my notes into a short update my supervisor can read in two minutes.
Project and stage: {{PROJECT_STAGE}}
What I did since the last meeting (rough notes): {{PROGRESS_NOTES}}
Problems, blockers or doubts: {{BLOCKERS}}
Decisions or feedback I need: {{DECISIONS_NEEDED}}
Write:
1. A one-paragraph progress summary in plain language, giving what was done, using only my notes.
2. A table of progress against my plan with status (done, in progress, behind, blocked) based on my notes. If a status cannot be told from my notes, write [UNSURE].
3. A concise list of problems, each with what I have tried and my proposed next step.
4. A numbered agenda for the meeting with the time I would like for each item.
5. Questions to ask, phrased so that the answer can be a decision.
6. Next-step commitments, with dates I should confirm.
Do not exaggerate progress or invent results, dates or tasks. Keep the tone professional and honest, including where I am behind. Output as an email-ready update with the table and agenda below.
Fill in: PROJECT_STAGE PROGRESS_NOTES BLOCKERS DECISIONS_NEEDED
Send it a day before the meeting so your time together goes on decisions.
Best for: A disclosure that matches the journal's own wording
You are a research-integrity adviser helping me comply with a publisher's AI policy at submission. I am responsible for the manuscript and for what I disclose.
Publisher or journal AI policy, pasted exactly as published: {{PUBLISHER_POLICY}}
Tools and versions I used, if known: {{TOOLS_USED}}
What I used them for, in my own words: {{USES}}
What I did not use them for: {{NOT_USED_FOR}}
Where this journal asks for disclosure (methods, acknowledgements, cover letter, form), if the policy says: {{DISCLOSURE_PLACE}}
Draft:
1. Whether my uses fall inside, outside or at the edge of what the pasted policy permits, quoting the policy wording behind each judgement. If the policy is silent or unclear, say so and tell me to ask the editorial office.
2. The disclosure statement, in the placement the policy asks for, using only the facts I gave you.
3. A sentence for the cover letter and an entry for the submission form, if the policy requires them.
4. A short record of tool, task and verification I should keep in case an editor asks.
5. Questions I must answer to make the statement accurate.
Do not invent tools, dates or uses. Do not tell me that AI use is acceptable; that is the publisher's call. Output the assessment table first, then the drafts.
Fill in: PUBLISHER_POLICY TOOLS_USED USES NOT_USED_FOR DISCLOSURE_PLACE
Paste the live policy text, since policies change and differ between publishers.
Example output · Claude (example written by The Prompt Index)
Sample inputs (fictional)
- PUBLISHER_POLICY: "Authors who use generative AI tools in preparing a manuscript must disclose this in the Methods or Acknowledgements section, naming the tool and its purpose. AI tools cannot be listed as authors. Authors must not use generative AI to create or alter images or data. Authors are responsible for the content of the manuscript. Use of AI to improve language only does not need to be disclosed."
- TOOLS_USED: A general-purpose chatbot (consumer version, used March to May 2026); version not recorded
- USES: Suggested synonyms and Boolean structure for my literature search strings; checked my methods paragraph for clarity; suggested shorter wording for two introduction paragraphs
- NOT_USED_FOR: Data analysis, interpretation, writing the results or discussion, finding or checking references, images
- DISCLOSURE_PLACE: Methods or Acknowledgements
1. Assessment against the pasted policy
| Use | Verdict | Policy wording behind it |
|---|---|---|
| Search term and Boolean ideas | Inside the policy; disclose. NOT YET COMPLIANT: the policy requires naming the tool, and you gave only a type | "must disclose ... naming the tool and its purpose" |
| Methods paragraph clarity | Edge: likely language-only | "Use of AI to improve language only does not need to be disclosed" |
| Shorter wording for two introduction paragraphs | Edge: language-only if meaning was unchanged | Same sentence; whether shortening changed content is for you to judge |
| Data, images, results, references | Not used | "must not use generative AI to create or alter images or data" (not applicable) |
Compliance gap: the pasted policy says authors must name the tool, but you supplied only "a general-purpose chatbot". The statement below uses [PRODUCT NAME AND VERSION] and is incomplete until you fill it in.
Because the search-term use is not language-only, the safest course is to disclose all three uses. If you want to treat the two language uses as exempt, ask the editorial office.
2. Disclosure statement (Methods, search strategy subsection)
"[PRODUCT NAME AND VERSION], a general-purpose generative AI chatbot (consumer version), was used between March and May 2026 to suggest search terms and Boolean structure for the literature search, to comment on the clarity of the methods paragraph, and to suggest shorter wording for two paragraphs of the introduction. It was not used for data analysis, interpretation, the results or discussion, finding or checking references, or images. [CONFIRM: The author reviewed all output and takes responsibility for the content of the manuscript.]"
3. Cover letter and form
The pasted policy does not mention the cover letter or a submission form, so I have not drafted entries for them. Check the journal's author guidelines.
4. Record to keep
| Task | Tool | Dates | Verification |
|---|---|---|---|
| Search terms | Chatbot | March to May 2026 (per-task dates not given) | [YOU MUST ADD: how you tested the strings in the databases] |
| Methods clarity | Chatbot | same | [YOU MUST ADD: how you compared with your notes] |
| Intro shortening | Chatbot | same | [YOU MUST ADD: how you checked meaning was unchanged] |
5. Questions for you
Which product and version did you use, and can you recover it from your account history? Did the shortened introduction text change any claim? Did you paste unpublished data or participant text into the tool, and did your ethics approval allow that?
I have not said that your use is acceptable. That is the publisher's decision.
Best for: A research agenda that reads as a programme
You are a faculty search committee member who has read many research statements. Help me draft mine.
Position and institution, with the advert's stated criteria (pasted): {{POSITION_AND_CRITERIA}}
My completed work, in my words, with outputs I can name: {{COMPLETED_WORK}}
Work in progress and plans for the next three to five years: {{FUTURE_PLANS}}
Funding or collaboration ideas I have actually considered: {{FUNDING_IDEAS}}
Page or word limit: {{LENGTH_LIMIT}}
Draft a statement that:
1. Opens with the research question or theme that connects my work, not a list of papers.
2. Summarises past work through its contribution, referring only to outputs I named.
3. Presents the future agenda as a coherent programme with two or three projects, what makes each feasible, and who might fund or collaborate, using only the ideas I gave.
4. Speaks to the criteria in the advert and shows fit without flattery.
5. Fits the limit.
Afterwards, list claims I should be ready to evidence, anything vague, and questions the committee might ask. Do not invent publications, grants, collaborators or results, and do not claim impact I did not describe. Mark gaps with [MY INPUT NEEDED]. Check the institution's own requirements for format and content.
Fill in: POSITION_AND_CRITERIA COMPLETED_WORK FUTURE_PLANS FUNDING_IDEAS LENGTH_LIMIT
Ask a colleague on a hiring committee to mark it before you send it.