IMED Research Publications puts this together as a recommended companion for everyone who reviews for IJPR, IJAR, and IJIMS — a distilled, working version of the guidance published by COPE, the ICMJE, and WAME, not a replacement for it. We offer it as guidance rather than a rulebook, because it's the same standard we hold ourselves to: IMED Research Publications and each of our journals follow COPE's Core Practices and the ICMJE Recommendations as a matter of policy across our own editorial process, and we ask reviewers to meet us there.
Peer review runs on trust, and that trust starts before you open the manuscript — at the moment you decide whether to accept the invitation at all. COPE's guidelines frame this as the reviewer's first real ethical decision.
Strip away the procedural detail and both COPE and the ICMJE converge on the same four obligations. They're simple to state and easy to breach without noticing.
A submitted manuscript is the authors' private property until it's published. Its existence, contents, and your comments on it aren't yours to share.
Financial ties, personal relationships, institutional overlap, or plain academic rivalry can all bias a review — even unintentionally.
Delay has a real cost — it holds up a decision the authors are waiting on and, for time-sensitive findings, the wider field too.
Bias creeps in around nationality, institution, seniority, language fluency, gender, or a prior disagreement with the authors.
WAME's description of a good review is a useful yardstick: professional, honest, courteous, prompt, and constructive. Here's how that translates into structure.
Two or three sentences on what the paper claims to do and how it tries to do it. This tells the editor — and the authors — that you actually engaged with the work, not just the flaws.
Major comments address whether the conclusions are supported by the methods and data. Minor comments cover clarity, presentation, and smaller fixes. Mixing the two buries the issues that actually matter for the editorial decision.
Note anything that raises questions about research conduct — consent, approvals, data handling, undisclosed conflicts — separately from the scientific critique, so it doesn't get lost.
Most systems let you write privately to the editor. Use it for genuine concerns you can't put to the authors directly (like suspected misconduct) — not as a place to soften what you actually think in the main review.
A recommendation (accept, revise, reject) is useful context, but the journal is under no obligation to follow it — editorial decisions weigh other things too, like fit for the journal and priorities you may not see.
This is one of the fastest-moving parts of reviewer policy right now, and three organizations have each put out guidance in the last two years: COPE, the ICMJE, and WAME. The wording differs slightly between them, but the substance lines up. Below is the practical version, with each point traced back to where it actually comes from — check your own journal's instructions too, since practice still varies by publisher.
Don't assume AI assistance is fine by default. Reviewers are expected to follow the journal's stated AI policy, or request permission before using any AI tool to help with a review.
ICMJE, Use of AI by Reviewers; WAME Recommendation 4, Chatbots & Generative AI (2023)Uploading a manuscript to a chatbot or writing assistant is a disclosure to a third party. WAME's board has been specific about why this matters: most chatbots retain whatever you type into them, including manuscript text, and that content can resurface in later responses to other users. If the journal hasn't confirmed confidentiality is preserved, don't paste the manuscript in.
WAME Recommendation 4, Chatbots & Generative AI (2023); ICMJE, Responsibilities in the Submission and Peer-Review ProcessIf a journal permits AI assistance and you use it, say so, and explain how. Don't present AI-drafted commentary as entirely your own independent judgment — both the ICMJE and WAME treat this as a matter of accountability, not just courtesy.
ICMJE, Use of AI by Reviewers; WAME Recommendation 4, Chatbots & Generative AI (2023)AI output can sound confident while being wrong, incomplete, or subtly biased. The scientific assessment has to remain yours — treat AI as a drafting aid at most, never as the reviewer.
ICMJE, Use of AI by ReviewersThat's a different situation from your own AI use, and it's not yours to adjudicate. COPE's guidance on a real case like this is consistent with its broader stance on suspected problems: note what you noticed — odd phrasing, an inconsistent voice, anything that reads generated rather than written — factually, in your confidential comments to the editor. Don't run the text through a public AI-detection tool yourself and treat the result as proof; detection tools are unreliable, and doing so can itself create a confidentiality problem. Let the editor investigate and decide.
COPE, "Reviewer suspects use of AI by an author" (case); COPE Ethical Guidelines for Peer Reviewers, v2 (2017)You're not expected to run a forensic investigation, but reviewers are often the first person to notice something off. If you do, don't try to resolve it yourself — flag it to the editor. COPE publishes flowcharts for exactly these situations.
| What you might notice | Why it matters |
|---|---|
| Text that reads familiar | Passages that closely echo another paper you know, or an oddly uneven writing style within one manuscript, can indicate unattributed copying. |
| Numbers that don't add up | Sample sizes, statistics, or figures that shift between the text, tables, and abstract are worth a second look — sometimes innocent, sometimes not. |
| Images that look reused or altered | Duplicated panels, suspicious splice lines, or unusually clean backgrounds in blots and micrographs are common flags in image integrity checks. |
| A paper that feels familiar overall | Substantial overlap with a paper you've already seen (by the same authors, elsewhere) may signal duplicate submission or salami-slicing of one study into several papers. |
| Missing ethical approvals | Human or animal research without a stated ethics approval, consent statement, or trial registration number is a gap worth raising directly. |
| A reviewer request that feels engineered | If you're asked to review because an author "suggested" you and something about the arrangement feels off, mention it to the editorial office. |
Primary sources, decision aids, and reference tools worth bookmarking — not just this page's summary of them.
The full source guideline this page is built on — models of peer review, reviewer responsibilities, and how to conduct a review.
publicationethics.org → COPE · FlowchartA decision flowchart for the exact moment covered in Section 01 above — useful the next time an invitation lands in your inbox.
publicationethics.org → ICMJE · RecommendationsThe ICMJE's own reviewer, editor, and author responsibilities — the backbone of biomedical journal policy worldwide.
icmje.org → ICMJE · AI PolicyThe short, specific section this page's AI guidance (Section 04) is drawn from.
icmje.org → WAME · RecommendationsWAME's five recommendations, including Recommendation 4 on reviewers disclosing chatbot use — and why pasting a manuscript into one breaches confidentiality.
wame.org → COPE · Focus CollectionCOPE's ongoing, regularly updated collection of AI-related guidance, case discussions, and webinars — useful when a journal's own policy is silent.
publicationethics.org → WAME · RecommendationsWAME's broader ethics framework, including its description of what a professional, constructive review looks like.
wame.org → EQUATOR NetworkLook up the right reporting checklist for the study type you're reviewing — CONSORT for trials, PRISMA for reviews, STROBE for observational studies, and more.
equator-network.org → COPE · DiscussionA candid look at copyright, credit, and confidentiality in reviewer reports — relevant if you're wondering what you can and can't reuse or share.
publicationethics.org → ORCIDMany journals can push a verified record of your review to your ORCID profile (with your permission) — a way to make otherwise invisible work count.
orcid.org →The whole page, compressed to what you actually need mid-review.
Everything above is a working paraphrase of these documents. Read them directly for anything you plan to cite or rely on formally.