Peer review is the unpaid, often invisible labor that holds scientific publishing together. Roughly 13 million reviews are written each year worldwide, the vast majority by working researchers who do the work in their evenings and weekends. The ethical norms that govern this process determine whether the scientific literature deserves the trust placed in it. In 2026, several of those norms are being tested by reviewer fatigue, generative AI, and a long-overdue debate about transparency. This guide walks through the responsibilities of both reviewers and authors, with the practical decisions you actually face in mind.
The Reviewer's Core Responsibilities
The Committee on Publication Ethics (COPE), the standards body most journals follow, lays out the reviewer's job in a short but demanding list. A reviewer should:
- Decline if not qualified. If you are not familiar with the methods or the subfield, the right answer is to decline and recommend a colleague who is. Reviewing outside your competence is a disservice to the authors and the journal.
- Disclose conflicts of interest. This includes recent collaboration with the authors (typically a 3- to 5-year window), shared institutional affiliation, financial stakes in the outcome, and personal relationships. Borderline cases should be disclosed and the editor allowed to decide.
- Respect confidentiality. A submitted manuscript is confidential. You may not share it with colleagues, post excerpts on social media, use it as training data for an AI model, or cite it in your own work before publication. Confidentiality is absolute.
- Be timely. If you accept a review, you owe the authors a response within the journal's stated window — usually two to four weeks. Sitting on a manuscript for months is one of the worst things a reviewer can do.
- Be constructive and specific. A useful review explains what is wrong, why it matters, and what would fix it. Vague rejections ("this needs more work") help no one. Personal attacks on the authors are never acceptable.
The Author's Side of the Bargain
Authors have ethical responsibilities too, and meeting them is the single most effective way to receive a fair review:
- Submit to one journal at a time. Simultaneous submission is considered misconduct by every major publisher. If a journal doesn't acknowledge your submission within a reasonable window, withdraw formally before submitting elsewhere.
- Disclose all relevant conflicts. Funding sources, commercial relationships, and prior versions of the work (including preprints) all belong in the cover letter and the manuscript itself.
- Make your data and code available. Most reputable journals now require this for any quantitative paper. Provide a working link, not a "data available on request" placeholder.
- Respond to reviewers in good faith. A thoughtful response letter that addresses each reviewer point — including the ones you disagree with — dramatically increases your chances of acceptance. Brushing off a reviewer's concern almost always backfires.
Single-Blind, Double-Blind, and Open Review
Three review models coexist in 2026, and the choice has real consequences for fairness:
- Single-blind. The reviewer knows who the authors are, but the authors don't know who the reviewers are. This is still the most common model. Its weakness is that it allows unconscious bias against authors from less prestigious institutions or underrepresented backgrounds.
- Double-blind. Neither side knows the other's identity. Studies have shown this measurably reduces gender and institutional bias in acceptance decisions. The drawback is that genuinely anonymizing a paper is hard — self-citations and writing style often give away the author.
- Open review. Both sides are identified, and in some journals (notably eLife and the BMJ family) the review reports are published alongside the article. Open review reduces some of the worst behavior and produces a much richer scientific record, but it makes some reviewers less willing to be sharply critical.
The Reviewer Fatigue Crisis
The number of submissions has roughly tripled since 2010, but the pool of qualified reviewers has not. Editors now report needing to send out four or five invitations on average to secure a single accepted reviewer. The result is delayed reviews, overworked specialists, and a real risk that the system will break under its own weight.
Several responses are being tried: payment for reviewers (controversial because it changes the incentives), mandatory review quotas tied to publication ("you may submit one paper for every two you review"), and explicit recognition of review work in tenure files. None has emerged as a clean solution, but the problem is taken seriously enough that we're likely to see structural change in the next five years.
AI in the Review Loop
Generative AI is now in the review process whether anyone admits it or not. The ethical lines being drawn in 2026 are roughly: it is acceptable to use AI to check your review for grammar and clarity; it is unacceptable to upload the manuscript to a public AI tool, because that violates confidentiality; it is unacceptable to have the AI write the substantive review for you; and it is misconduct to use AI to fabricate reviewer reports as a journal editor.
If you are an editor, ask reviewers explicitly to disclose any AI use. If you are a reviewer, when in doubt, don't.
The Practical Takeaway
Ethical peer review is not a code to memorize. It is a habit of treating other researchers' work with the same care you would want for your own. If you can imagine being on the receiving end of a review and feeling that it was fair, specific, and timely — even if the answer was "reject" — you are doing it right.