Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    How peer review works

    Peer review is the critical assessment of a submitted manuscript by experts who are usually not on the journal's staff; the ICMJE says its real value is widely debated, but it gives papers a fair hearing and often improves them.[1][2] Editors choose reviewers and make the final decision, and journals differ widely in how many reviewers they use and whether names are disclosed.[3][4][5] Newer models publish the reviews alongside the paper, or review a study's design before any data are collected.[6][7]

    Editor reviewedUpdated Explainer19 claims · 100% primaryScience

    What peer review is

    The International Committee of Medical Journal Editors (ICMJE) defines peer review as the critical assessment of manuscripts submitted to journals by experts who are usually not part of the editorial staff.[1] It is not a guarantee. The ICMJE says the actual value of peer review is widely debated, but that it gives manuscripts a fair hearing, helps editors decide what suits their journals and often improves the quality of reporting.[2]

    The same idea is used by funders. The US National Science Foundation and the National Institutes of Health both send grant proposals to outside experts before deciding what to fund; see how research is funded.[8][9]

    Step by step at a journal

    1. Submission and editor screening. Journals are not obliged to send every manuscript out for review, nor to follow reviewers’ recommendations. The editor is ultimately responsible for what the journal publishes.[3]
    2. Choosing reviewers. At PLOS ONE, for example, editors choose reviewers for their expertise in the manuscript’s research area.[5]
    3. Review. Reviewers assess the work against the journal’s criteria. PLOS ONE’s criteria focus on rigour: analyses performed to a high technical standard and conclusions supported by the data. It also considers negative and null results.[10]
    4. Decision. The editor makes the decision and is not bound by the reviewers’ recommendations.[3]

    Journals vary at every step: how many and which manuscripts they send for review, how many reviewers they use and whether review is open or anonymized. The ICMJE therefore says each journal should publish a description of its process.[4]

    Anonymized or open

    Under single-anonymized review, the default at PLOS ONE, reviewers know the authors’ names but authors do not learn who reviewed them, although reviewers may choose to sign their reports.[11] Journals differ in whether review is open or anonymized.[4]

    Reviewers’ obligations

    ICMJE guidance sets out reviewers’ duties. They should keep manuscripts confidential, must not use the authors’ ideas before publication, and should declare relationships that could bias their opinion and step aside if there is a conflict.[12] Reviewers must get the journal’s permission before using AI tools in a review, because uploading a manuscript may breach confidentiality.[13]

    New models: preprints, public reviews and Registered Reports

    Some journals, PLOS ONE among them, consider manuscripts that have already been posted on preprint servers such as bioRxiv or arXiv.[14]

    eLife has gone further. It makes no accept-or-reject decision after peer review: every article it reviews is published as a Reviewed Preprint, with the public reviews, an author response and an eLife Assessment.[6] The assessment rates a paper’s significance on a scale from landmark to useful and the strength of its evidence from exceptional to inadequate.[15] Editors who are active researchers decide which submissions to review, and the $3,750 publication fee is waived for authors who cannot afford it.[16]

    Registered Reports move review earlier. Peer review happens before data are collected, and protocols that pass are provisionally accepted for publication if the authors follow the registered method.[7] The Center for Open Science, which promotes the format, says it removes questionable practices such as low statistical power, selective reporting and publication bias.[17] Why that matters is covered in the reproducibility crisis.

    When review misses problems

    Publication after review is not the final word. Papers later found to be seriously flawed can be retracted, and journals are expected to label them clearly; see why scientific papers are retracted.[18][19]

    Next in the crash course

    Questions readers ask

    What is peer review?

    The International Committee of Medical Journal Editors (ICMJE) defines it as the critical assessment of manuscripts submitted to journals by experts who are usually not part of the editorial staff.[1]

    Who decides whether a paper is published?

    The editor. Journals are not obliged to send papers for review or to follow reviewers' recommendations; the editor is ultimately responsible for what the journal publishes.[3]

    Do authors know who reviewed their paper?

    It depends on the journal. Under single-anonymized review, used by default at PLOS ONE, reviewers know the authors' names but authors do not learn the reviewers' names unless reviewers choose to sign.[11][4]

    Does peer review prove a paper is right?

    No. The ICMJE says the actual value of peer review is widely debated. PLOS ONE, for example, asks reviewers whether the analyses are technically sound and the conclusions supported by the data, not whether the result is important.[2][10]

    Can reviewers use AI tools?

    ICMJE guidance says reviewers must get the journal's permission first, because uploading a manuscript may breach confidentiality.[13]

    What is a Registered Report?

    A format in which peer review happens before data are collected. Protocols that pass review are provisionally accepted for publication if the authors follow the registered method.[7]

    Sources

    · 19 claims from 6 publishers

    Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.

    1. [1]

      The ICMJE defines peer review as the critical assessment of manuscripts submitted to journals by experts who are usually not part of the editorial staff. confirmedas of 2026-11-08

    2. [2]

      The ICMJE says the actual value of peer review is widely debated, but that it gives manuscripts a fair hearing, helps editors decide what suits their journals and often improves the quality of reporting. confirmedas of 2026-11-08

    3. [3]

      Journals are not obliged to send manuscripts for review or to follow reviewers' recommendations; the editor is ultimately responsible for what the journal publishes. confirmedas of 2026-11-08

    4. [4]

      Journals differ in how many and which manuscripts they send for review, how many reviewers they use and whether review is open or anonymized, so the ICMJE says each should publish a description of its process. confirmedas of 2026-11-08

    5. [5]

      At PLOS ONE, editors choose reviewers based on their expertise in the manuscript's research area. confirmedas of 2026-11-08

    6. [6]

      eLife has no accept-or-reject decision after peer review; every article it reviews is published as a Reviewed Preprint with public reviews, an author response and an eLife Assessment. confirmedas of 2026-11-08

    7. [7]

      In the Registered Reports format, peer review happens before data are collected, and protocols that pass are provisionally accepted for publication if the authors follow the registered method. confirmedas of 2026-11-08

    8. [8]

      NSF program officers send proposals to outside expert reviewers, either ad hoc, on a panel or both, who judge them on two National Science Board criteria, intellectual merit and broader impacts; NSF aims to tell applicants within six months. confirmedas of 2026-11-08

    9. [9]

      NIH evaluates all research funding applications that meet basic requirements through two-stage peer review, in which panels of mostly non-federal scientists rate applications for scientific merit and the advisory councils of NIH's 24 research institutes and centers then make funding recommendations; an institute or center director makes the final decision. confirmedas of 2026-04-17

    10. [10]

      PLOS ONE's publication criteria focus on rigour - analyses performed to a high technical standard and conclusions supported by the data - and it considers negative and null results. confirmedas of 2026-11-08

    11. [11]

      PLOS ONE uses single-anonymized review by default, in which reviewers know the authors' names but authors do not learn the reviewers' names, and reviewers may choose to sign their reviews. confirmedas of 2026-11-08

    12. [12]

      Under ICMJE guidance reviewers should keep manuscripts confidential, must not use authors' ideas before publication, and should declare relationships that could bias them and recuse themselves if there is a conflict. confirmedas of 2026-11-08

    13. [13]

      The ICMJE says reviewers must get the journal's permission before using AI tools in a review, because uploading a manuscript may breach confidentiality. confirmedas of 2026-11-08

    14. [14]

      PLOS ONE considers manuscripts that have already been posted on preprint servers such as bioRxiv or arXiv. confirmedas of 2026-11-08

    15. [15]

      eLife Assessments rate a paper's significance on a scale from landmark to useful and the strength of its evidence on a scale from exceptional to inadequate. confirmedas of 2026-11-08

    16. [16]

      At eLife, editors who are active researchers decide which submissions to peer review, and the publication fee of $3,750 is waived for authors who cannot afford it. confirmedas of 2026-11-08

    17. [17]

      The Center for Open Science says Registered Reports remove questionable research practices such as low statistical power, selective reporting and publication bias. confirmedas of 2026-11-08· interpretation

    18. [18]

      The ICMJE says errors serious enough to invalidate a paper's results may require retraction, and honest errors that change the conclusions can be handled by retraction and republication of a corrected paper. confirmedas of 2026-11-08

    19. [19]

      The ICMJE says retracted articles should stay publicly available but be clearly labelled as retracted, and the retraction notice should explain why. confirmedas of 2026-11-08

    Revision history (1)
    1. Page created.

    Created Nov 8, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: May 8, 2027.

    Cite this page

    "How peer review works." ContentLora, updated Nov 8, 2026. https://contentlora.com/explain/how-peer-review-works

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.