Learning & Research

Summarize a Research Paper Critically

Summarizes a paper's question, method, findings and limits, and flags what the evidence does and does not support.

Designed for general-purpose AI assistantsPublished November 2026

Customize

Fill in the fields and the prompt updates. What you type stays in this page and is never saved or sent.

Preview

Summarize the research paper below so I can decide whether to read it in full and how much to trust it.

My background in this field: general reader
Why I am reading it: {{purpose}}
Specific questions I want answered, if any: none

Paper text or abstract plus relevant sections:
{{paper}}

Cover these points, in plain language and in this order:
1. The question: what the authors set out to find out, and why it matters.
2. The method: study design, who or what was studied, sample size, how outcomes were measured and over what period. If the text does not say, write "not stated".
3. The findings: the main results, with the numbers exactly as written, including effect sizes, confidence intervals or p-values when reported. Do not round or restate them.
4. How strong the evidence is: the study type, sample, controls or comparison group, and what that means for how far the findings can be generalized. Distinguish correlation from causation where the design leaves it open.
5. Limitations: the ones the authors state and any further ones that are visible from the method I gave you. Mark yours as [my observation].
6. What the authors claim versus what the data shown supports: flag any conclusion that goes further than the results.
7. Terms I may not know, explained in a sentence each.
8. A verdict on whether it is worth reading in full, and which sections to read first for my purpose.

Rules: only use the text I pasted. Do not describe methods, results or citations that are not in it. If I gave only the abstract, say clearly that the method and limitations cannot be properly assessed from the abstract alone. Do not state that a paper has been peer reviewed or replicated unless the text says so.

2 required fields are empty and will stay as a {{placeholder}} in the copied text.

How to use it

  1. Paste the abstract and the methods and results sections if you can, not just the abstract. Abstracts rarely show a study's weaknesses.
  2. Say why you are reading the paper. A paper can be excellent and still irrelevant to your question.
  3. Check key numbers against the original tables before quoting them anywhere.
  4. Use the follow-up question 'what would change your assessment of this paper?' to find what to look for in the full text.

Example output

Illustrative only. Real answers depend on your input and the assistant you use.

Question: Does a four-week sleep extension program improve reaction time in college athletes?

Method: Randomized trial, 62 athletes at one university, assigned to sleep extension or usual habits. Reaction time was tested at the start and end. Duration: four weeks.

Findings: Reaction time improved by 8.1 ms in the intervention group and 1.2 ms in the control group (reported difference 6.9 ms, 95% CI 2.1 to 11.7).

Strength of evidence: Moderate for the short term. Randomized, but a small, single-site sample of athletes, so results may not apply to other groups. Sleep was self-reported.

Limitations: authors note the short duration. [my observation] No blinding is described and participants knew which group they were in.

Claims vs data: the abstract says the program "improves performance", but only reaction time was measured, not match performance.

Terms: "CI" (confidence interval) is a range of values consistent with the data.

Verdict: worth reading the methods and Table 2 if you work with athletes; skip the discussion section.

Tips

When the assistant says something is "not stated", take it as a lead. Open the full paper and look, because sometimes the information is in an appendix.

A summary is not a replacement for reading a paper you plan to cite. Use it to decide where to spend your attention.

Paste a second paper and ask for a comparison of designs only if both are in the conversation. Otherwise the assistant may recall details of papers from memory that are wrong.

If you want an assistant to search a literature database itself instead of working from text you paste, the MCP Directory lists a PubMed server. Check its listing for what it covers, and still verify every citation it returns.

  • Learning & Research

    Analyze a Long Document With Citations

    Answers your specific questions about a long document with quoted evidence and section references, and says what the document does not cover.

    Customizable fields
    5 fields
  • Learning & Research

    Compare Options in a Decision Matrix

    Builds a weighted decision matrix from your options and criteria, shows how sensitive the result is to your weights and flags missing facts.

    Customizable fields
    5 fields
  • Writing & Editing

    Write an Executive Summary

    Condenses a long report or proposal into a one-page executive summary that leads with the decision needed and the evidence for it.

    Customizable fields
    6 fields