Research · Reading & synthesis
The Best Prompt for Summarizing Research
The best prompt for summarizing research tells the model who the summary is for and holds it to the document's own numbers and level of certainty. This prompt checks that the pasted text is complete, then produces a fixed structure: bottom line, method, key findings with their location in the text, where claims outrun evidence, limitations, and implications for the reader. Anything from outside the document is fenced off in its own section.
Research Paper and Report Summary Prompt
Summarizes a paper or report for a named reader: bottom line, method, key findings with exact figures, limits of the evidence, and what it means for them.
- Works with
- ChatGPT, Claude, Gemini
- Difficulty
- Beginner
- Complexity
- Multi-part reply
- Variables
- 6
- Version
- 1.0
Customize this prompt
Fill in what you know. Anything left blank stays as a [bracketed placeholder]. Nothing you type leaves your browser.
6 variables · fields marked * matter most
# Role You are a careful analyst who summarizes research for people who will not read the original. Accuracy to the document matters more than fluency. # What I need - Document type: [Document type] - Who will read the summary: [Who will read the summary] - What they will use it for: [What they will use it for] - Length: [Length] - Give extra attention to: [Give extra attention to] # Document <document> [Document text] </document> # Before you write Read the whole document. Then check three things and report them in one short note at the top: whether the text appears complete or cut off; whether tables, figures or appendices are referred to but missing from what I pasted; and what kind of evidence this is (for example experiment, survey, review of other studies, modeling, opinion). # Summary structure 1. Bottom line - two or three sentences: what the document set out to answer and what it concludes, at the level of confidence the authors themselves use. 2. How they got there - data, sample, method and time period, in plain language pitched at the reader above. 3. Key findings - up to six, ordered by importance to the reader's purpose. Reproduce each number exactly as written, with its unit and comparison group, and say where it appears (section, table or page as shown in the text). 4. Claims versus evidence - points where the conclusions go further than the results shown, or where a finding rests on a single analysis. 5. Limitations - split into "stated by the authors" and "my observations". Keep the two lists separate. 6. What it means for this reader - implications for the stated purpose, and what the document cannot tell them. 7. Terms - brief definitions of any jargon the reader would need, using the document's own definitions where given. # Constraints - Use only the document. If you add outside context, put it in a final section headed "Outside the document" and keep it to three points at most. - Do not round, convert or recalculate figures unless you show the original next to yours. - Preserve hedging. Do not turn "may", "suggests" or "is associated with" into firm or causal statements. - If the document does not say something the reader would expect (sample size, funding, dates), write "Not stated" instead of inferring it. - No praise or criticism of the writing. No filler introduction. # Output format The completeness note first, then the seven numbered headings. Findings as a numbered list; everything else in short paragraphs.
Use it when you are
- Briefing a manager or client on a long industry or policy report
- Deciding whether a journal article is worth reading in full
- Preparing notes on a paper for a journal club, seminar or team meeting
- Explaining a technical study to a non-specialist audience
- Building a consistent set of summaries across many documents
Why this prompt works
A summary is written for someone
The same paper needs a different summary for a clinician, a budget holder and a student. Naming the reader and what they will do with it lets the model choose which findings lead, how much method to explain and which terms need defining, instead of compressing every section equally.
The completeness check catches a quiet failure
Pasted PDFs often lose tables, figures or the final pages. A model will summarize whatever it receives without comment unless asked. Reporting whether the text looks complete, and which referenced tables are absent, tells you up front whether the key numbers were ever in view.
Numbers stay as written and carry a location
Rounding, unit conversion and dropped comparison groups are where summaries drift from their source. Requiring exact figures with the section or table they came from makes each finding quick to verify and discourages numbers that sound plausible but appear nowhere in the document.
Author limitations and model observations are kept apart
When the two are blended, a reader cannot tell what the authors conceded from what the model inferred. Two labeled lists let you quote the first with confidence and treat the second as prompts for your own judgment, to be confirmed against the paper before you repeat them.
Hedging is protected
Compression tends to strip qualifiers, so "associated with" becomes "causes" and "may" becomes "does". The explicit instruction to preserve the authors' level of certainty, together with the claims-versus-evidence section, addresses one of the most common ways a summary ends up overstating what a study found.
How to use it
- Paste the full text, including tables if you can. If you only have the abstract, say so in the document type field and expect a thinner result.
- Describe the reader by role and background, and write the purpose as a decision or task.
- Read the completeness note before anything else. If it reports missing tables or a cut-off, fix the paste and rerun.
- Spot-check two or three key findings against the locations given.
- Ask follow-ups for depth, for example: "Explain finding 2 and the analysis behind it for someone new to regression."
- Remove or verify the "Outside the document" section before sharing the summary with anyone else.
AI model compatibility
| Model | Status | Notes |
|---|---|---|
| ChatGPT | Expected to work | If you upload a PDF instead of pasting text, read the completeness note carefully: tables and figures do not always survive extraction. |
| Claude | Expected to work | Anthropic's guidance for long documents suggests placing the document near the top of the prompt and the instructions after it. If your paper is very long, move the <document> block above "What I need". |
| Gemini | Expected to work | Keep the numbered structure and the output format section as written; Google's prompting guidance favors explicit instructions and a stated response format. |
“Expected to work” means the prompt uses plain, model-independent structure and has been reviewed for that — not that it has been run on that model. A model is marked Tested only when a test record is published below.
Worked example
Scenario: An operations lead needs a short briefing on a published evaluation of a nurse scheduling change before a planning meeting.
# Role You are a careful analyst who summarizes research for people who will not read the original. Accuracy to the document matters more than fluency. # What I need - Document type: Journal article - Who will read the summary: A hospital operations manager with no statistics background - What they will use it for: Deciding whether to propose a pilot of self-scheduling on two wards - Length: About 400 words - Give extra attention to: Effects on overtime and staff turnover # Document <document> [Full text of the article pasted here, including the methods section and results tables.] </document> # Before you write Read the whole document. Then check three things and report them in one short note at the top: whether the text appears complete or cut off; whether tables, figures or appendices are referred to but missing from what I pasted; and what kind of evidence this is (for example experiment, survey, review of other studies, modeling, opinion). # Summary structure 1. Bottom line - two or three sentences: what the document set out to answer and what it concludes, at the level of confidence the authors themselves use. 2. How they got there - data, sample, method and time period, in plain language pitched at the reader above. 3. Key findings - up to six, ordered by importance to the reader's purpose. Reproduce each number exactly as written, with its unit and comparison group, and say where it appears (section, table or page as shown in the text). 4. Claims versus evidence - points where the conclusions go further than the results shown, or where a finding rests on a single analysis. 5. Limitations - split into "stated by the authors" and "my observations". Keep the two lists separate. 6. What it means for this reader - implications for the stated purpose, and what the document cannot tell them. 7. Terms - brief definitions of any jargon the reader would need, using the document's own definitions where given. # Constraints - Use only the document. If you add outside context, put it in a final section headed "Outside the document" and keep it to three points at most. - Do not round, convert or recalculate figures unless you show the original next to yours. - Preserve hedging. Do not turn "may", "suggests" or "is associated with" into firm or causal statements. - If the document does not say something the reader would expect (sample size, funding, dates), write "Not stated" instead of inferring it. - No praise or criticism of the writing. No filler introduction. # Output format The completeness note first, then the seven numbered headings. Findings as a numbered list; everything else in short paragraphs.
What a good response should contain
- A completeness note stating whether the results tables came through and what type of study it is
- A bottom line that mirrors the caution in the article's own conclusion
- Findings on overtime and turnover listed first, each with the figure as printed and the table or section it comes from
- A limitations section with separate lists for the authors' stated limits and the model's own observations, such as how far a single-site result transfers
- A plain statement of what the study cannot tell the manager about their own wards
This is an illustrative scenario showing how to fill in the prompt and what to look for. It is not a recorded model output. Recorded outputs will be published here once this prompt has a test record.
Testing and editorial review
Status: Untested. Reviewed for structure and clarity by the editorial team. No recorded model test yet.
No test record has been published for this prompt yet. When it is tested, this section will show the model and version, the date, the inputs, the actual output and notes against our seven-point rubric. We do not publish numerical scores.
| Author | TheBestPromptFor Editorial Team — The editorial team writes, reviews and maintains the prompts and guides on TheBestPromptFor. Prompts are drafted with AI assistance and reviewed against published editorial standards. |
|---|---|
| Published | November 15, 2026 |
| Last reviewed | November 15, 2026 |
| Last tested | Not yet tested |
| Version history |
|
Limitations and things to check
- The model can misread tables, especially when columns are scrambled by copying from a PDF. Check key figures against the original.
- A summary is no substitute for reading the methods when a decision carries real consequences.
- The model cannot assess whether the study has been corrected, retracted or contradicted by later work unless you tell it.
- Its own observations about limitations are informed guesses, not peer review.
Alternative prompts
Research
Source-Bound Literature Review Prompt
Turns the papers, abstracts and notes you paste in into an evidence table, themes and a synthesized draft, with every citation traceable to a source you supplied.
- ChatGPT
- Claude
- Gemini
Research
Show-Your-Working Data Analysis Prompt
Analyzes a pasted table or dataset step by step: data check, calculations with working shown, findings tied to those calculations, and what the data cannot support.
- ChatGPT
- Claude
- Gemini
Productivity
Meeting Notes to Decisions and Actions Prompt
Converts a transcript or rough notes into decisions, owned action items and open questions, without assigning anything the meeting did not.
- ChatGPT
- Claude
- Gemini
Education
Study Guide from Your Notes Prompt
Builds a study guide from your own notes or readings: topic map, core content, commonly confused ideas, self-test questions with separate answers, and a revision schedule.
- ChatGPT
- Claude
- Gemini
Related resources
Guide
How to Write a Good AI Prompt
The five parts of a complete prompt (goal, context, task, constraints, output format), with a worked rewrite and a pre-send checklist.
Guide
How to Improve AI Responses
A diagnostic approach to weak AI answers: name the fault, trace it to the gap in your prompt, then follow up precisely or rewrite and start again.
Frequently asked questions
Can I use this on a paper I have not read at all?
Yes, as a way to decide whether and where to read. Do not cite or act on a finding you have only seen in the summary. The locations attached to each finding are there so you can go straight to the relevant table or section and confirm it.
What if the document is too long to paste?
Summarize it in parts: run the prompt on each major section with the same reader and purpose, then paste those section summaries into a new conversation and ask for a combined version in the same seven-part structure. Say in the final summary that it was built from partial summaries.
Why does the prompt restrict outside knowledge?
So you can tell what the document says from what the model believes. Background from the model may be accurate, outdated or wrong, and mixing it into the findings makes the summary impossible to check against its source. A separate, capped section keeps it visible and optional.
Sources and references
- Prompt engineering overview — Anthropic
- Prompt design strategies — Google AI for Developers
Was this prompt useful?
No account needed. Do not include personal information.