Expertise
Name the claims and methods you can assess. Tell the editor which decision-relevant area needs another reviewer rather than pretending broad competence.
Sources and scope · Guide reviewed September 22, 2026
- Coverage
- Invitation fit, conflicts, confidentiality, claim-evidence assessment, issue triage, constructive comments, AI boundaries, and editor handoff
- Source basis
- COPE, ICMJE, Nature, BMJ, PLOS, and Elsevier primary reviewer guidance
Prepared by the Manusights editorial team. Publisher and institutional instructions take precedence.
Reviewer evidence desk
Review the evidence.
Not the authors.
A useful review connects a manuscript observation to the claim it affects, explains the consequence, and asks for the smallest in-scope resolution. This record keeps that chain visible.
8 primary sources6 decision gates8 final audit testsEdition 2026.09.22-review-1
Five-page reviewer evidence record
Peer review decision record
Record invitation fit, policy boundaries, central claims, evidence, issue priority, author-facing comments, editor-only concerns, and the final consistency audit. Licensed CC BY 4.0 for the original Manusights framework.
- Edition
- 2026.09.22-review-1
- Format
- License
- CC BY 4.0
- Evidence reviewed
- September 22, 2026
Before opening the manuscript
An invitation is a boundary decision, not an obligation
Accept only when your expertise, independence, time, and the journal's process support a fair review. Disclose uncertainty instead of resolving it privately. COPE and ICMJE both require reviewers to declare relevant conflicts, protect confidentiality, and respond within the agreed time. Source A Source B
Conflict
Record financial and nonfinancial relationships, competition, collaboration, institutional proximity, and any circumstance that could bias the evaluation.
Capacity
Confirm the deadline and expected review stage. If the work cannot be completed carefully and on time, decline or request a feasible change.
Policy
Save the exact journal instructions governing identity, confidentiality, co-reviewing, transparent review, AI, data access, and report publication.
Original Manusights framework
Six gates connect the invitation to the editorial handoff
Treat each gate as a record another editor or reviewer could inspect. The report should not depend on an undocumented impression.
- 1
Invitation boundary
Can I review this manuscript fairly, competently, confidentially, and on time?
Record
Name the expertise you can cover, the areas you cannot cover, the deadline, every relationship or activity that could bias the review, and the journal's decision on any disclosed uncertainty.
Evidence of completion
The invitation, current reviewer instructions, conflict disclosure or recusal record, accepted deadline, and any editor clarification are preserved before manuscript assessment begins.
- 2
Policy boundary
What may I do with the manuscript, tools, notes, and assistance?
Record
Capture the journal's confidentiality, co-reviewing, identity, transparent-review, AI, data-access, note-retention, and communication rules. Treat silence as an unresolved policy question, not permission.
Evidence of completion
A dated policy record identifies whether colleague or trainee involvement, AI assistance, report publication, reviewer signing, and external processing are allowed and what disclosure is required.
- 3
Claim-evidence map
What are the manuscript's central claims, and what evidence is offered for each one?
Record
For each decision-relevant claim, cite the manuscript location, linked methods, result, uncertainty, boundary condition, and the authors' stated limitation. Separate what is reported from what you infer.
Evidence of completion
Every substantive concern points to a stable manuscript locator and the affected claim rather than a general impression of the paper or authors.
- 4
Validity assessment
Where does the evidence support, narrow, or fail to establish the claim?
Record
Assess the relevant design, sampling, methods, analysis, results, interpretation, reporting, ethics, and reproducibility dimensions within your expertise and the journal's criteria.
Evidence of completion
Each issue states the observation, its consequence for a named claim or decision criterion, the evidence location, and the smallest in-scope information or change needed to resolve it.
- 5
Comment construction
Can the author and editor understand the issue, its importance, and the bounded request?
Record
Build comments as observation, consequence, request, and scope. Order them by decision relevance, number them, use neutral language, and distinguish author-facing improvement from confidential editorial information.
Evidence of completion
The report is specific, evidence-linked, professionally phrased, free of personal judgment, and does not demand a different study merely because it would be interesting.
- 6
Consistency and handoff
Do the report, confidential comments, and recommendation tell one evidence-consistent story?
Record
Compare the recommendation with the stated journal criteria and decision-relevant issues. Route conflicts, confidentiality problems, or integrity concerns to the editor without investigating them yourself.
Evidence of completion
The final audit records unresolved uncertainty, recommendation rationale, editor-only matters, required disclosures, submission date, and secure deletion or retention action under journal policy.
Claim-evidence assessment
Review the inferential chain, not a generic paper outline
PLOS describes the reviewer's core jobs as determining whether the evidence supports the authors' claims and helping the editor apply the journal's criteria. Start with the central claim, then follow the evidence that carries it. Source F
Claim and contribution
Inspect: The central question, claim, contribution, comparison, and population or setting to which the claim applies.
Ask: Is the claimed contribution stated precisely, and does the manuscript distinguish what the study shows from what it assumes or extrapolates?
Boundary: Novelty and journal significance are editorial criteria. Apply the named journal's standard instead of inventing a universal threshold.
Design and sampling
Inspect: The design's ability to answer the stated question, selection process, comparator, timing, sample, inclusion and exclusion rules, and sources of bias.
Ask: Does the design support the type and scope of inference being made, and are important threats to validity visible?
Boundary: Do not require a different design simply because it would answer a broader question. Identify the claim that the present design cannot support.
Methods and reproducibility
Inspect: Interventions, exposures, procedures, materials, measures, protocols, deviations, data provenance, code or analytic workflow, and reporting needed to understand or repeat the work.
Ask: Could a qualified reader understand what was done, why it was done, and where material uncertainty or discretion entered?
Boundary: A reviewer cannot certify reproducibility from prose alone. State which information is missing and what conclusion depends on it.
Analysis and uncertainty
Inspect: Outcome definitions, estimands or target quantities, analytic choices, assumptions, missing data, multiplicity, sensitivity analyses, effect sizes, uncertainty, and model diagnostics relevant to the claim.
Ask: Do the analyses match the design and question, and are uncertainty and analytical limitations represented in the conclusions?
Boundary: If the analysis is outside your competence, tell the editor. Do not provide statistical certification beyond your expertise.
Results and presentation
Inspect: Consistency among text, tables, figures, supplements, denominators, units, labels, reported outcomes, and the analysis described in the methods.
Ask: Can the result supporting each central claim be located and interpreted without contradictions or missing context?
Boundary: Prioritize evidence and interpretability over copyediting. A list of language edits is not a substitute for scientific review.
Interpretation and limits
Inspect: Whether conclusions track the results, alternative explanations, limitations, generalizability, causal language, practical importance, and relationship to prior evidence.
Ask: Would a narrower claim, additional evidence, clearer limitation, or different framing repair the mismatch?
Boundary: Do not ask authors to cite your work or add speculative discussion unless it is directly necessary and objectively justified.
Issue triage
Classify by decision effect, not reviewer effort
Journal labels such as major, minor, accept, revise, or reject vary. Use the journal's categories in the form. Before that translation, classify what the issue actually does.
Decision-critical
Use when: The issue changes whether a central claim meets a named journal criterion or can be supported by the reported study.
Write: Name the claim, evidence location, consequence, and smallest in-scope resolution. Explain why the editor needs the issue for the decision.
Avoid: Do not label an issue major merely because it is lengthy, technically interesting, or would improve a future study.
Revision-relevant
Use when: The issue affects transparency, interpretation, reproducibility, or reader use but does not by itself overturn the central claim as currently bounded.
Write: Specify the ambiguity or omission and the exact clarification, analysis, display, or limitation needed.
Avoid: Do not bury a decision-critical issue in a long list of presentational requests.
Editorial-only
Use when: The information concerns your conflict, review limitation, confidentiality, suspected integrity problem, policy question, or another matter the journal must handle privately.
Write: Describe the observation and uncertainty without accusing, investigating, or making a finding. Follow the journal's confidential channel.
Avoid: Do not hide scientific feedback from authors merely because it supports your recommendation.
Optional suggestion
Use when: The idea could improve clarity or usefulness but is not required to support the present claim or meet the journal's criteria.
Write: Label it optional and keep it proportionate.
Avoid: Do not convert preference, citation promotion, or a next-study idea into a publication condition.
Confidentiality and AI
No tool receives the manuscript by default
A submitted manuscript is confidential material, not ordinary text for a convenience service. ICMJE requires reviewers to follow journal AI policy or request permission, avoid systems where confidentiality cannot be assured, disclose permitted use, and remain responsible for accuracy. Nature prohibits manuscript upload to generative AI during peer review. Source C Source D
Policy
Retrieve the exact journal rule and any editor clarification. A publisher's general page may not answer the journal's form or system boundary.
Data path
Determine what content leaves the journal system, who receives it, whether it is retained or reused, and whether confidentiality can actually be assured.
Accountability
If a use is permitted, disclose it as required and independently verify every output. The reviewer, not the tool, owns the report and its errors.
Submission audit
Eight independent tests before the report leaves your control
A polished review can still fail if it breaches confidentiality, overstates an integrity concern, expands scope, or contradicts its own recommendation.
Fit test
Pass: Expertise coverage, limitations, conflicts, and deadline are explicit and any uncertainty has been disclosed to the editor.
Repair: Pause before review and ask the editor to narrow the assignment, clarify a conflict, extend the deadline, or invite another reviewer.
Confidentiality test
Pass: The manuscript, notes, data, report, and communications remain within journal-authorized people, systems, and uses.
Repair: Do not share with a colleague, trainee, service, or AI system until the exact policy and required permission are established.
Locator test
Pass: Every substantive comment identifies the manuscript evidence and the claim or criterion it affects.
Repair: Replace vague reactions with a stable locator, observation, consequence, and bounded request.
Inference test
Pass: The report distinguishes reported facts, reviewer interpretation, unresolved uncertainty, and suspected integrity concerns.
Repair: Remove accusations and overclaims; route unresolved integrity concerns to the editor for journal process.
Scope test
Pass: Required work is necessary to assess or support the present study under the named journal criteria.
Repair: Relabel an interesting next study as optional or remove it rather than making publication contingent on scope expansion.
Fairness test
Pass: The report addresses the manuscript rather than the authors, acknowledges material strengths, and uses professional, neutral language.
Repair: Remove sarcasm, speculation about motives or competence, identity-based assumptions, and unsupported comparisons.
Consistency test
Pass: Issue priority, author-facing report, confidential comments, and recommendation are mutually consistent and tied to journal criteria.
Repair: Reconcile contradictions and move author-actionable scientific feedback into the author-facing report.
Handoff test
Pass: Disclosures, contributors, AI use if permitted, unresolved limits, editor-only concerns, submission date, and record disposition are documented.
Repair: Complete required disclosures and follow the journal's retention or deletion instructions before closing the review.
Evidence and maintenance
Source register and policy limits
The web guide, PDF, and source record share one edition. Cross-journal ethics guidance and publisher examples define a defensible floor, but the assigned journal's current policy remains controlling.
- ACOPE Ethical Guidelines for Peer Reviewers
Committee on Publication Ethics / Before, during, and after review; preparing a report. Supports: Accepting only with suitable expertise and time, declaring conflicts, confidentiality, authorized co-reviewing, constructive and objective reporting, avoiding personal use, consistent confidential comments, and routing suspected misconduct to the journal rather than investigating it.
Boundary: COPE provides cross-journal ethical guidance. The exact journal policy and review model still control the assignment.
- BResponsibilities in the Submission and Peer-Review Process
International Committee of Medical Journal Editors / Section II.C.2.c, Peer Reviewers. Supports: Confidentiality, prompt responses and agreed timing, acknowledgment of trainee or colleague assistance, constructive and polite comments, conflict disclosure and recusal, secure deletion, and the editor's responsibility for the final decision.
Boundary: ICMJE recommendations focus on medical journals and do not replace a journal's current reviewer instructions.
- CUse of AI by Reviewers
International Committee of Medical Journal Editors / AI use by reviewers. Supports: Following the journal's AI policy or requesting permission, protecting confidential manuscripts from systems where confidentiality cannot be assured, disclosing permitted use, and retaining human responsibility for validity.
Boundary: Permission, confidentiality, disclosure, and allowed use vary by journal and system. This source is not blanket authorization to use AI.
- DPeer Review Policy
Nature / Reviewer conduct, confidentiality, co-reviewing, and AI. Supports: Strict confidentiality, prior editor consultation before involving a colleague, identification of contributing colleagues, timely communication, human reviewer accountability, and restrictions on uploading manuscripts to generative AI tools.
Boundary: This is a Nature policy. Other journals and publishers can use different review models and disclosure rules.
- EResources for Reviewers
BMJ / Reviewer responsibilities and peer-review terms. Supports: The editor's final-decision role, manuscript confidentiality, disclosure of co-reviewers and conflicts, professional comments, and confidential routing of suspected misconduct to the editor.
Boundary: BMJ journals can have journal-specific forms, criteria, open-review practices, and terms that must be checked for the assignment.
- FHow to Read a Manuscript as a Peer Reviewer
PLOS / Reviewer role and manuscript assessment. Supports: Identifying central claims, assessing whether evidence supports them, using journal publication criteria, applying subject expertise, and prioritizing research evaluation over copyediting.
Boundary: The page is practical publisher guidance. Apply the exact PLOS journal criteria only to a PLOS review and verify another journal's criteria separately.
- GHow to Write a Peer Review
PLOS / Reviewer report structure and effective feedback. Supports: Organized reports, concrete evidence and examples, numbered and located comments, professional and respectful language, proportional requests, and avoiding out-of-scope experiments or self-promotion.
Boundary: The suggested structure is useful but not universal. Follow the assigned journal's form, labels, and confidential-comment rules.
- HHow to Conduct a Review
Elsevier / Before, during, and after review; reviewer ethics. Supports: Confidential handling, restrictions on AI upload, human control of permitted supportive tools, open-minded assessment, avoidance of ad hominem comments, proportionate language feedback, and communication with the editor.
Boundary: Publisher-level guidance does not establish the criteria or form for every Elsevier journal. Check the exact journal instructions.
Scope: A decision and evidence record for invited reviewers assessing a scholarly manuscript. It does not replace the journal's current reviewer instructions, subject or method expertise, editorial judgment, research-integrity procedures, statistical review, ethics review, or permission to share or process a confidential manuscript. Recheck all primary sources after a material publisher, editorial-ethics, confidentiality, AI, or peer-review policy change and no later than 90 days after this review. At the time of use, verify the exact journal criteria, review model, confidentiality terms, AI policy, conflict process, recommendation labels, and editor instructions.
Common questions
Policy first, then judgment
Can I ask a trainee or colleague to help with a review?
Only after checking the journal's current policy and obtaining any required editor permission. COPE, ICMJE, Nature, and BMJ guidance all require transparency about assistance, but the exact permission and disclosure process varies. The additional reviewer must also protect confidentiality and disclose conflicts.
Can I upload a manuscript to an AI tool to help write the review?
Do not assume that you can. ICMJE says reviewers must follow the journal's stated AI policy or request permission, protect confidential material from systems where confidentiality cannot be assured, disclose permitted use, and remain responsible for validity. Nature prohibits reviewers from uploading manuscripts to generative AI tools. Verify the exact journal and tool terms before any processing.
How long should a peer review be?
There is no defensible universal word count. The report should be long enough to identify the central claims, explain decision-relevant issues with manuscript locators, and make bounded requests, but short enough that priorities remain clear. Follow the journal's form and remove repetition, copyediting lists, and optional next-study ideas.
What should I do if I suspect fabrication, plagiarism, or another integrity problem?
Record the specific observation and raise it confidentially with the editor under the journal's process. Do not accuse the authors, contact them directly, or investigate the matter yourself. COPE guidance assigns investigation and follow-up to the journal, not the reviewer.
Cite this resource
Citation formats
Free to use and cite, with attribution. See the reuse policy for what you may republish.
- APA
- Manusights. (2026). Peer review decision record. Edition 2026.09.22-review-1. https://manusights.com/resources/how-to-peer-review
- MLA
- Manusights. "Peer Review Decision Record." Manusights, 22 Sept. 2026, https://manusights.com/resources/how-to-peer-review.
- Vancouver
- Manusights. Peer review decision record [Internet]. 2026 Sep 22. Available from: https://manusights.com/resources/how-to-peer-review
Comment architecture
Turn concern into a bounded, evidence-linked request
A comment is useful when the author can locate the issue and the editor can see why it matters. PLOS recommends specific examples, numbered comments, professional language, and requests that stay within the present study. Source G
Observation
What exactly is reported, absent, inconsistent, or unclear? Cite the page, line, figure, table, supplement, or section.
Consequence
Which claim, interpretation, reproducibility need, reader decision, or journal criterion is affected, and how?
Request
What is the smallest specific clarification, evidence, analysis, correction, limitation, or editorial action that would resolve the issue?
Scope
Is the request necessary for this study and within the journal's criteria, or is it an optional suggestion or future-work idea?
Worked structure
Observation: The primary outcome is defined differently in the methods and Figure 2 legend. Consequence: The reader cannot determine which result supports the stated primary conclusion. Request: Reconcile the definition across the methods, analysis, figure, and interpretation, and report which definition was prespecified. Scope: This asks for internal consistency and provenance, not a new study.