You can tell whether something is peer reviewed by looking for the journal's stated peer-review policy, the article's publication metadata, and the presence of a formal review history. A journal name, a DOI, or a PubMed listing alone does not prove review happened. You need evidence that an independent editor sent the work to outside experts before acceptance, and that the work was revised in response.
Peer review is a gatekeeping process, not a quality guarantee. This explainer covers how review actually works, what the evidence looks like, where the signal fails, and how to verify a claim of review in about five minutes.
Mechanism: what peer review actually verifies
Peer review is a workflow, not a label. A manuscript arrives at a journal and is screened by an editor for scope and basic integrity. If it clears that desk screen, the editor invites two or three independent researchers who work in the same narrow specialty. Those reviewers read the methods, check the statistics, and assess whether the claims match the data. They return comments and a recommendation: accept, revise, or reject.
The critical detail is what review cannot do. Reviewers usually see only the manuscript and its supplementary files. They rarely see raw data, participant consent forms, or the underlying code. They are unpaid and time-limited, and they are selected by the editor, not randomly. So review is a plausibility check by informed peers, not a replication and not an audit.
Review also has structural variants that change what the label means. Single-blind review hides reviewers from authors. Double-blind review hides both directions, which modestly reduces bias against authors from less prestigious institutions. Open review publishes reviewer names or the full report alongside the paper. Post-publication review moves evaluation after publication. Each variant is legitimate, but "peer reviewed" describes none of them specifically, which is why the journal's stated policy matters more than the word itself.
One more mechanism matters for detection: the revision record. Almost every accepted paper went through at least one round of revision. Many publishers now deposit that history — submission date, revision dates, reviewer reports — as part of the article's public metadata. That record is far stronger evidence of real review than any badge.
Quantified context: how much review exists, and how fast
The scale is larger than most readers assume. Estimates put the global volume of peer-reviewed research at roughly two to three million articles per year, spread across tens of thousands of journals. Review is the binding constraint on that pipeline, and it is done largely as unpaid labor by a relatively small pool of active researchers.
Turnaround varies by field and by journal tier. Desk rejection typically happens within days to two weeks. Full external review commonly takes four to twelve weeks for a first decision, and a revise-and-resubmit cycle adds another two to six months. High-selectivity journals in biology and medicine often run longer, with total time from submission to acceptance frequently exceeding six months.
The practical implication: a paper claiming very rapid acceptance is not automatically suspect, but very short intervals combined with very fast revision rounds deserve a closer look at the review record. Likewise, a journal that advertises acceptance in under two weeks across its whole catalog is signaling a screening process, not a review process.
Edge cases: where the label bends or breaks
The label gets murky at several well-documented boundaries. Conference proceedings in computer science are genuinely peer reviewed, often more rigorously than journal articles, but they are not indexed as journal publications. Book chapters may be edited but not externally reviewed. Preprints are posted before review and are explicitly unreviewed, even when the authors are eminent. Registered reports are reviewed before the data are even collected, which is a stronger design but confuses readers who expect review to follow results.
A counterintuitive professional caveat: acceptance after review does not mean the reviewers endorsed the conclusions. Many reviewers recommend revision on the grounds that the methods are sound while explicitly disagreeing with the interpretation. The paper is published because it passed methodological scrutiny, not because experts agreed with it.
Predatory publishing is the sharpest edge case. These journals charge authors fees, promise fast acceptance, and claim peer review that is perfunctory or nonexistent. A 2023 analysis estimated that a substantial share of biomedical articles appear in journals with no meaningful review, though exact counts are disputed and change yearly. The warning signs are behavioral, not cosmetic.
The journal solicits you by email within days of a preprint appearing. Acceptance arrives within days of submission with no substantive reviewer comments. The journal's scope spans unrelated fields such as dentistry and aerospace. The publisher lists a physical address that resolves to a residential or virtual office. Reviewer reports, when shared, contain no specific line-level criticism.
Misconceptions: signals that do not prove review
Several widely used signals are weaker than they look. A DOI proves only that a persistent identifier was registered; registration requires no review. Indexing in PubMed or Scopus means the journal met that database's inclusion criteria at some point, which usually includes a review policy on paper but not continuous verification of every article. An impact factor measures citation frequency, not review rigor, and it is calculated at the journal level, not the article level.
The strongest misconception is that "peer reviewed" means "true." Landmark failures have passed review, including papers later retracted for fabricated data and clinical findings later overturned by larger trials. Review catches some errors and misses others, and its reliability varies by field. Treat the label as one evidence tier among several, and weight replication and independent confirmation more heavily.
Practical actions: a five-minute verification sequence
Work through these in order and stop when you have enough evidence.
Open the journal's own website and find its peer-review policy page. It should name the review model, state how many reviewers are typical, and describe the decision authority. Check the article page for dates: received, revised, accepted, published. A gap of several weeks to months between received and accepted is consistent with real review. Look for a published review history, transparent peer-review file, or reviewer reports link. Confirm the journal appears in a recognized index such as MEDLINE, Scopus, or Web of Science. Search the publisher's name plus "predatory" and check whether it appears on maintained watchlists. Read one reviewer comment if available and check whether it addresses specific methods or statistics.
Use this comparison to weigh what each signal actually establishes.
| Signal | What it proves | What it does not prove |
|---|---|---|
| Journal peer-review policy page | The journal claims a review process | That this article went through it |
| Received-to-accepted dates | Time existed for external review | That review was substantive |
| Published reviewer reports | Review occurred and was specific | That reviewers endorsed conclusions |
| MEDLINE or Scopus indexing | Journal met database criteria | Article-level review quality |
| DOI | Persistent identifier registered | Any review at all |
| Impact factor | Journal citation frequency | Rigor of any single paper |
| Preprint server listing | Early dissemination | Peer review, ever |
A typical review timeline looks like this, though fields vary widely.
| Stage | Typical duration | What it indicates |
|---|---|---|
| Submission to desk decision | 3–14 days | Editorial screening only |
| Desk decision to first external decision | 4–12 weeks | Active external review |
| First decision to revised submission | 4–12 weeks | Author revision work |
| Revised submission to acceptance | 2–8 weeks | Re-review and editor decision |
| Acceptance to publication | 1–8 weeks | Production, not review |
| Preprint posting to journal acceptance | 6–18 months | Full cycle including review |
Reconcile the thresholds before you judge. Fast desk rejection within two weeks is normal and says nothing bad. Fast acceptance within two weeks, paired with no revision dates, is the pattern that warrants suspicion. The gap between received and accepted is the single most informative date pair.
Limitations: what verification cannot establish
No publicly available signal tells you whether the reviewers were competent, whether the editor overrode them, or whether the authors responded honestly to criticism. Transparent review files are increasingly common but still a minority of published output. Watchlists of predatory journals are incomplete and lag behind new launches, so absence from a watchlist is not clearance.
Field norms also confound comparison. A computer science conference paper and a clinical trial report carry different review standards and different failure modes, and comparing them by any single metric misleads. Where evidence about a specific article's review is missing, the honest conclusion is that review status is unverified, not that it was faked.
The most reliable posture is layered: treat peer review as one filter, weigh replication and independent citation more heavily, and read the methods section yourself when the claim matters. That habit outperforms any badge.