ChatGPT prompt for scientific writing

Draft a [SECTION TYPE] for [AUDIENCE OR VENUE] about [RESEARCH QUESTION OR TOPIC]. Use only the evidence in [SOURCE MATERIAL AND CITATIONS]. The study or analysis context is [METHODS AND RESULTS], and the required style or word limit is [STYLE AND LENGTH].

Before drafting, make a claim-evidence map: list each proposed substantive claim, the exact supporting result or source I supplied, and its appropriate strength of wording. If evidence is absent, mark the claim unsupported rather than filling the gap.

Write the section in a publication-ready scientific style. Use a logical sequence appropriate to the section type: background should move from established knowledge to the unresolved gap; methods should be reproducible enough to assess design and analysis; results should report findings without interpretation; discussion should distinguish findings, interpretation, limitations, and implications. Define abbreviations on first use. Preserve numerical values, units, uncertainty measures, sample sizes, and statistical notation exactly as supplied. Use cautious language for observational, exploratory, or underpowered findings.

Return: 1) the draft; 2) a short “evidence and limits” note listing unsupported claims, missing methodological details, and overstatements to avoid; and 3) suggested locations for citations using only the citations I supplied. Do not create sources, data, participant details, methods, p-values, or causal conclusions. Self-check that every factual claim is traceable to my material and that the conclusion does not exceed the design. Ask up to 3 clarifying questions only if a required input is missing.

Write a [WORD COUNT]-word Methods section from [STUDY PROTOCOL], [ANALYSIS PLAN], and [ETHICS DETAILS] for [VENUE]. Organize it as design and setting, participants, variables or intervention, data collection, outcomes, statistical analysis, and ethics where applicable. Preserve all numbers, dates, instruments, software, thresholds, and missing-data rules exactly as provided. Identify any information needed for reproducibility but absent from my material. Do not invent approvals, recruitment procedures, sample-size calculations, or analytic choices. End with a checklist of missing details; ask up to 3 questions only if required inputs are missing.

Turn [TABLES AND FIGURES] and [ANALYSIS NOTES] into a [WORD COUNT]-word Results section for [AUDIENCE]. Report participant flow, baseline characteristics, prespecified primary outcomes, then secondary and sensitivity analyses in that order when supplied. State effect estimates with units, confidence intervals, and p-values exactly as given; do not interpret clinical importance or explain causes. Reference each table or figure once at the relevant point. Flag discrepancies between totals and subgroup counts. Do not calculate or infer unreported statistics. Ask up to 3 questions only if a required table label or outcome definition is missing.

Write a plain-language research summary for [PUBLIC AUDIENCE] using [STUDY DETAILS] and [RESULTS]. Explain the question, who took part, what was done, main findings, uncertainty, limitations, and what the results do and do not mean. Use short sentences and define necessary technical terms in context. Produce a title, a 250-word summary, and five factual takeaways. Do not promise a health outcome, imply individual medical advice, or add facts beyond the supplied material. Flag claims that need researcher review. Ask up to 3 questions only if a required input is missing.