Peer Review Policy

1. Overview

1.1 · What this policy does. XSci is a preprint platform with a paid certification service on top of it. This policy sets out what review on XSci is, what it decides, and what it does not decide.

1.2 · Depositing is free and unconditional. Depositing a manuscript on XSci is never conditional on any purchase. A deposit is dated and is assigned a DOI — a permanent, citable identifier that resolves to that record indefinitely. Every subsequent revision is deposited as a new version with its own DOI. Depositing is free to read, carries no embargo, and does not depend on the outcome of any review.

1.3 · Review is a service purchased separately, and its outcome does not affect whether a manuscript is available or citable. Because an assessment is bought against a deposited version identified by its DOI, the DOI necessarily exists before an assessment can begin. A manuscript that is refused certification remains a preprint on the platform, dated and citable exactly as before, and the refusal leaves no public trace.

1.4 · Human review requires a deposit first. Human review of a manuscript may be purchased only once that manuscript has been deposited. XSci does not assess unposted work.

1.5 · Must and may. Where this policy uses must, the obligation is binding on the person or party named. Where it uses may, a genuine discretion is intended.

2. Core principles

2.1 Quality and integrity

2.1.1 · What review supports. Review supports quality control and scholarly integrity. It is not a guarantee that a manuscript's findings are correct.

2.1.2 · What a decision rests on. A review decision must rest on the merits of the version assessed and on the published Certification Standard, and on nothing else.

2.1.3 · No member of XSci staff may ask a reviewer to change a decision, and none has the authority to do so. XSci is paid by the person whose work is assessed; this rule is what keeps that payment away from the judgement.

2.1.4 · Neither the assessment fee nor the reviewer payment varies with the decision. Throughout this policy, the assessment fee is what an author pays XSci for an assessment, and the reviewer payment is what XSci pays the reviewer who carries it out; the two are separate amounts and neither is calculated from the other. No part of a reviewer payment depends on the outcome, and a refusal is a complete assessment delivered in full, not a failure of the service.

2.2 What is fixed and what may vary

2.2.1 · Some features of review on XSci are fixed and are not varied by service type, price, or workflow. They are 2.2.2 to 2.2.6.

2.2.2 · One decision, and it is binary. A version is either certified or it is not. There is no score, grade, rating, band, star or ranking, and none may be introduced.

2.2.3 · One version, one purchased review, one reviewer. A purchaser may buy exactly one review of any given version. A second opinion on the same version is not available at any price. Where more than one reviewer is ever used, the number is fixed before any report exists, reviewers do not see one another's reports, the record states the split in words and never rounds it up, and one certifying reviewer is sufficient. This rule has no exception. Where an appeal succeeds on ground 2 the certification is revoked and the author is refunded, and the refund is the whole of the remedy: no replacement reviewer is appointed and the version is not assessed again (clause 7.3).

2.2.4 · One author response per assessment (clause 6).

2.2.5 · The two prices, and what the lower one buys (clause 3.2).

2.2.6 · The appeal route (clause 7.3).

2.2.7 · What may vary is the presentation of a report, the tooling used to support it, and the delivery timetable within the range stated in clause 3.5.

2.3 Transparency

2.3.1 · The criteria are published. The full criteria, findings and refusal conditions applied by a reviewer are published in the XSci Certification Standard, which is dated and versioned and is free to read without an account. Every reference to the Standard in this policy links to that page. A certification record names the version of the Standard applied to it, and links to it.

2.3.2 · What XSci publishes about itself. From the first certification sold, XSci will endeavour to publish at least annually: the number of assessments completed; the certification rate, computed over assessments, with the denominator stated; and the distribution of assessments per certified manuscript. Nothing is ever published about an individual reviewer.

2.3.3 · The figures are not independently auditable. These figures come from XSci's own records and no reader can independently audit them, because a refusal leaves no public trace. XSci states this alongside the figures rather than implying a publicly checkable denominator.

2.3.4 · A manuscript assessed more than once is not thereby suspect. Revising after a refusal and buying a further assessment is the service working as intended, and the published figures report it as ordinary information.

2.3.5 · Corrections and retractions are governed by the Retraction and Correction Policy (clause 7.1).

3. Review process

3.1 Initial deposit

3.1.1 · What a deposit is and is not. A manuscript deposited on XSci is published, dated and assigned its DOI. Depositing is not peer review, decides nothing about quality, and is not a review outcome of any kind — a deposited manuscript carries no assessment and implies none.

3.1.2 · Unlawful or unauthorised deposits. Where a deposit is unlawful, or was made without the authority of its authors, it is dealt with under the Terms of Service and, once published, under the Retraction and Correction Policy.

3.2 Review services and prices

3.2.1 · Two services. XSci offers AI Review and Human Review, distinguished in clause 4. Only Human Review can lead to certification.

3.2.2 · Two prices, and the disclosed price is the lower. When Human Review is purchased, the purchaser states whether the reviewer's report will be disclosed — that is, published alongside the manuscript record. The reduction is a discount given in exchange for publication of the report, not a surcharge for keeping it private.

3.2.3 · Automatic publication at the discounted price. A report bought at the discounted price is published automatically, 48 hours after the report is released to the author. The author takes no further step, and there is no stage at which the author confirms it. This is stated wherever the discounted price is offered.

3.2.4 · Withdrawing disclosure costs the difference between the two prices, is charged once only per assessment, and is calculated at the prices in force on the date of purchase, so a later change to the price list never reaches back into an assessment already bought.

3.2.5 · Opening a report bought at the higher price is free at any time, and is not refunded.

3.2.6 · There is no closing date on either election.

3.2.7 · A certification is always public, whether or not the report behind it is. Where a report is private, the certification record still states the version assessed, the date, the target venue it was measured against, the Standard version applied, and the number of reviewers who certified.

3.3 Reviewer assignment

3.3.1 · Where reviewers come from. Reviewers are compensated independent experts, invited from the authorship of closely related published work, and from among researchers who have published in the target venue named for the assessment, with a preference for the latter where a suitable reviewer is available.

3.3.2 · How they are matched. Reviewers are matched on subject-matter expertise and on the absence of any barred interest under clause 5.1. XSci may use internal tools to assist in identifying candidates, but reviewer assignment is never determined solely by an automated system, and no automated system certifies anything, alone or in combination.

3.3.3 · Single-blind, and no direct contact. Reviews are delivered under XSci's default single-blind arrangement: the reviewer's identity is not disclosed to the author. There is no direct contact between author and reviewer in either direction; all communication passes through the report and the response mechanism in clause 6.

3.4 Review criteria

3.4.1 · Both limbs must hold. Certification requires that both of 3.4.2 and 3.4.3 hold. They are independent, and both are judged by the reviewer.

3.4.2 · (a) No major error. A major error is a defect which, while it stands, means the manuscript's central claims are wrong, unsupported, or impossible to evaluate. A defect that ordinary revision would put right is not a major error.

3.4.3 · (b) A genuine prospect of publication, after ordinary revision, in the venue the authors are aiming at. The purchaser names up to three target journals when the assessment is bought; where several are named, the test is met if the manuscript would fit any one of them. Naming a target is optional. Where none is named, the target is the leading journals of the manuscript's field, and the reviewer must identify and name them in the report.

3.4.4 · The bar is relative to the target named. A sound manuscript aimed higher than it can reach is not certified, and the reviewer must recommend where the work would fit instead. That recommendation is advice and never a condition: a reviewer must not offer to reconsider if the author names a different target, and must not make certification conditional on any further purchase.

3.4.5 · Fit is a required section of every report, kept separate from the assessment domains below because the reviewer's judgement on fit, and the recommendation that follows from it, rest on different evidence.

3.4.6 · The nine domains and the four findings. Within limb 3.4.2, the report records, for each of the nine assessment domains set out in the Certification Standard and for each domain the manuscript engages, exactly one of four findings — no concern, concern noted but not material, material concern, or not applicable with reason — together with at least one sentence specific to that manuscript. These findings must not be summed, averaged, expressed as a fraction, converted to a number or a letter, or used to sort or rank anything, and are never displayed as the outcome. Not applicable describes the manuscript, not the reviewer: it is recorded only where the manuscript does not engage the domain at all, and it is never a mark against the work. A reviewer who cannot assess a domain the manuscript does engage must tell XSci rather than record it as not applicable.

3.4.7 · Underlying materials and the three steps. Where the manuscript's claims rest on underlying materials — whatever another competent person would need in order to check them, which depending on the field may be data, code, transcripts, field notes, source texts, images, instrument or archival records, derivations, protocols or specimens, no kind being treated as better evidence than another — the report records how far the reviewer personally got, in three steps: whether the materials are declared; whether they are reachable and as described; and whether the reported results follow from them.

3.4.8 · This is a record of how far one reader got, and nothing more. Whether the findings would hold up if someone collected new data is never assessed and never implied. A manuscript with no dataset and no code is not for that reason less well verified.

3.4.9 · The attestation. Every report carries the reviewer's attestation that they examined the full manuscript and the materials the report relies on, and is written to be understood by a competent researcher outside the immediate subfield.

3.5 Decisions

3.5.1 · Two outcomes, and no others. The reviewer records exactly one of:

(a) Certify. The version is accepted as an XSci Peer-Certified Preprint. (b) Do not certify, because — with each reason stated specifically. "Revise and I will judge then" is not an available outcome.

3.5.2 · Outright refusal conditions. Certification is refused outright, whatever the merits elsewhere, where any of the following is present: ethics, consent or regulatory approvals are absent or inconsistent with the work described; the methods are too incompletely described for the analysis to be evaluated; reported quantities cannot be reconciled with the stated design or sample and are not corrected; the central claims are unsupported by the evidence presented and are not narrowed; or a competing interest material to the findings was not disclosed.

3.5.3 · Integrity matters are referred, not decided. Where instead a reviewer encounters plagiarism, undisclosed duplicate publication, or signs that images or data have been manipulated, that is not a matter for the reviewer to decide. The reviewer must refer it to XSci and stop, and must not write the allegation into a report that may be published. While such a referral is open, the assessment shows as in progress with no public reason given, because a public integrity flag before anyone has made a finding would be unfair to the authors. The matter is then handled under clause 7.1.

3.5.4 · The completeness check. Every report is checked by XSci for completeness, not for its conclusion, before it reaches the author. Where a report is incomplete it is returned to the reviewer, up to three times, and the assessment is then reassigned to another reviewer. The certification decision is never a subject of this check.

3.5.5 · Delivery. Human Review is delivered within five to ten days of assignment in the ordinary case. This is a target rather than a guarantee, and where a report is not delivered at all the Refund Policy applies.

3.5.6 · The certification mark: one mark, no tiers.

3.5.7 · Two rules bind the mark wherever it appears, on XSci or anywhere else:

(a) It must never be displayed for a version that has not been certified, or for a version whose certification has been revoked. (b) It must never be shown with a number, tier, percentage, star or band, or described as a rating, score, grade or level of assurance.

3.5.8 · Two further rules govern how the mark is captioned: it must name the version certified, the date, and the target venue it was measured against, and must link to the certification record. On XSci these two are binding. Off XSci they are what we ask of you and recommend, because we can set the terms of display on our own pages but not on anyone else's; where you display the mark on a site of your own we ask you to caption it the same way.

3.5.9 · Nothing produced by any other XSci service may be called certification. The separate XSci Preprint and AI-reviewed labels are governed by clause 4.

4. AI review and human review

4.1 · The services differ in who is accountable for the report the author receives.

4.2 · AI Review. The report is generated by AI and is labelled as AI-generated. Authors choose among AI review services from more than one provider. Neutrality means three things: no editorial preference — we do not present our own service as the default or recommend it over another; the same commercial terms for every provider, including ours; and any published quality measure computed by a published method, applied identically to all.

4.3 · Human Review. A named expert is accountable for the report the author receives. The expert may use AI tools only as an aid in drafting or revising the report and remains responsible for its final content.

4.4 · No machine certifies anything. Certification requires the judgement of a compensated human expert under clauses 3.4 and 3.5, and an AI system cannot supply it, alone or in combination with another system.

4.5 · AI review output is not a verification. It is a generated narrative report, and is never displayed as though it had established a fact. AI outputs do not by themselves guarantee any outcome, and should be interpreted with appropriate care.

4.6 · Labelling. A manuscript that has been through AI review may carry the AI-reviewed label, and only once the author has deposited a revision following it — AI review run with no revision deposited changes nothing on the record. The label records that a service was run and asserts nothing about its adequacy or about the quality of the work. AI-reviewed may never be shortened to "reviewed" in interface text, marketing, or any badge image, and must be set so that the qualifier cannot be cropped away.

4.7 · Data. XSci does not train on an author's manuscript, or on anything supplied with it. Content published on the platform may be used to evaluate, research and improve XSci's review systems, including by training and fine-tuning XSci's own models, under the licence in the Terms of Service. Unpublished activity on the platform is used de-identified, may be used for training, and may be opted out of; confidential material — the manuscript and anything else supplied in order to obtain a service and not published — is never used this way at all. XSci may also use reviewer reports to improve its review systems, as described in the AI and Data Usage Disclosure and, for reviewers, the Reviewer Agreement. Whether a reviewer chose to see AI review output for a manuscript, and the reviewer's own use of AI in preparing the report, are recorded on the assessment record.

5. Reviewer expectations

5.0 · Where the binding form is. Reviewer obligations are set out in binding form in the Reviewer Agreement. The following are the requirements that bear directly on what a certification means, and are stated here so that authors and readers can rely on them.

5.1 Interests that bar, and interests that are declared

5.1.1 · Four bars. A reviewer must decline an assignment where any of the following applies:

(a) a prior role in this work — having advised on the manuscript, reviewed it for a journal, or taken part in producing or correcting any part of it. There is one exception: the reviewer of a certified version may assess a later version of the same manuscript; (b) a close personal relationship, or a dispute, with any author, of a kind a reasonable person would consider material; (c) a financial or commercial interest in the outcome, or in the subject matter to the extent that the outcome would affect it; (d) a current supervisory relationship in either direction with any author.

5.1.2 · Everything else is declared and is not a bar — including recent co-authorship, a shared institutional affiliation, competing work, academic rivalry, and strongly held prior public positions. The declaration goes on the assessment record, and XSci decides whether to proceed.

5.1.3 · A reviewer who discovers a barred interest after accepting an assignment must stop and disclose it, and bears no consequence for doing so.

5.2 Conduct

5.2.1 · A reviewer must:

(a) assess the version assigned against the published Certification Standard, and against nothing else; (b) state every reason for a refusal specifically, and record a finding and a manuscript-specific sentence for every domain the manuscript engages; (c) treat the report, the fact of the assessment, and any material not already public as confidential, and not transmit any of it to a third-party service not authorised by XSci; (d) refer an integrity matter under clause 3.5.3 rather than decide it.

5.2.2 · A reviewer must not:

(a) contact an author directly, or accept contact from one; (b) condition certification on a further purchase, or on the author naming a different target; (c) take credit for the assessment. Assessing a manuscript is paid work and is never authorship, contributorship, or acknowledgement.

6. Author rights and responsibilities

6.1 · What an author must do. Provide accurate deposit information; declare relevant competing interests and any overlapping or duplicate submission; declare at deposit whether a new version is a substantial revision of an earlier one; and, where the manuscript's claims rest on underlying materials, state what those materials are and where they can be reached.

6.2 · To deposit free of charge, and to have that deposit dated and given a DOI, whatever any review says.

6.3 · One response per assessment, after the decision. The report and the decision are released together, and the author is then entitled to one response. It is commentary recorded on the record beside the report: it does not reopen the assessment and does not change the decision, the report, or the version a certification names. A revised version is a new deposit and a new assessment, bought on the ordinary terms and at the ordinary assessment fee.

6.4 · The one-response limit applies to the assessment only, and never to an integrity inquiry. Where an integrity matter is raised about a manuscript, the author's answers to that inquiry are not the response described in 6.3 and are not counted against it. An author may make as many representations as the inquiry requires, until it is concluded, and having already responded to a reviewer's report takes nothing away from that. The two processes ask different questions and are kept separate; see the Retraction and Correction Policy.

6.5 · To a report meeting the standard of detail required by the Certification Standard, checked for completeness before delivery.

6.6 · To decide whether the report is published, on the terms in clause 3.2.

6.7 · To appeal, on the grounds and terms in clause 7.3.

6.8 · To revise, deposit a new version, and buy a further assessment at the ordinary assessment fee. This is the service working as intended and is not held against the manuscript.

6.9 · Which versions a certification covers. A certification is issued against the version the reviewer read, and the record names that version permanently. Where authors later deposit a further version, the certification stands and the mark carries forward, and authors are never asked to pay to certify a revision. A version the authors have declared a substantial revision says so beside the mark. Re-assessment may be purchased at any time at the ordinary assessment fee. A certification does not expire — the date attaches to the assessment, not to the mark. No claim is made, at any point, that a later version was read.

7. Post-publication issues, revocation, and appeals

7.1 Corrections, retractions, and integrity findings

7.1.1 · Where they are handled. Corrections, retractions, expressions of concern, and integrity matters referred under clause 3.5.3 are handled under the Retraction and Correction Policy. A DOI, once assigned, remains resolvable, and a corrected or retracted record remains accessible with its status shown.

7.2 Revoking or correcting a certification

7.2.1 · The only ground. A certification is revoked only on a finding that the record was materially false. It is never revoked because a reviewer changed their mind, because later work disagreed, or because the manuscript became controversial.

7.2.2 · The standards a revocation is decided against. A decision to revoke a certification, and any appeal against one, is taken in accordance with the guidance of the Committee on Publication Ethics (COPE) — which governs in the first instance — together with the recommendations of the International Committee of Medical Journal Editors (ICMJE) and the applicable Crossref metadata standards. Where a revocation is contested, it is those standards, and not the four appeal grounds in 7.3.4, against which the decision is measured. The same standards govern the corresponding decisions under the Retraction and Correction Policy, so that a single set of facts is judged one way wherever it arises.

7.2.3 · A revoked certification is never deleted. The record remains, shown as revoked, with the date and a reason.

7.2.4 · Errors are corrected; judgements are not. A factual error on a record is corrected with the original retained alongside it. A judgement is never corrected. A reviewer wishing to add something does so as a dated addition.

7.3 Appeals

7.3.1 · Two different appeals, not decided the same way.

7.3.2 · An appeal against a decision to certify or not to certify lies on the four grounds in 7.3.4, and the remedy is a refund.

7.3.3 · An appeal against a decision to revoke a certification does not lie on those grounds. It is decided against the standards named in 7.2.2 — COPE in the first instance, with ICMJE and the applicable Crossref metadata standards — by a person who took no part in the revocation decision. There is no fee. The remedy is that the revocation is set aside and the certification restored to the record, with the revocation notice corrected to show that it was set aside and the date; or the revocation stands, and the record is unchanged. No refund arises either way, because a revocation appeal is not a complaint about the assessment the author bought.

7.3.4 · The four grounds — appeals against a certification decision. Such an appeal lies on four grounds only:

(a) the Certification Procedure was not followed; (b) a reviewer had an undisclosed interest that would have barred them under clause 5.1; (c) the report contains an error of fact about the manuscript; (d) the report does not meet the standard of detail the Certification Procedure requires.

7.3.5 · Disagreement with the reviewer's judgement is not a ground, and that includes the judgement on fit.

7.3.6 · How it is decided, and the remedy. An appeal is decided by a person who took no part in the assessment. There is no fee. The remedy is a refund of the assessment fee. The report stands as written, the record stands, and the decision does not change — an appeal is not a route to certification. An author who wants a different decision revises, deposits a new version, and buys a further assessment.

7.3.7 · Where an appeal succeeds, the reviewer payment for that assessment is not due. Each of the four grounds is a finding that the assessment was not delivered as required, so the payment is withheld or, if already made, recovered under the Reviewer Agreement — after the reviewer has been told the ground and given an opportunity to answer it, with no limit on the exchanges and, where a question of fact cannot be resolved, an adjudication by XSci. This turns on conduct and completeness, never on the decision reached, and the author's refund does not depend on it: XSci owes the refund in full whatever the outcome for the reviewer.

7.3.8 · Where the successful ground is ground 7.3.4(b), and the manuscript was certified, two things follow together. The certification is revoked under clause 7.2, because a record which does not disclose an interest that barred the reviewer is materially false. And the author is refunded the assessment fee, because what they bought was an assessment by an eligible reviewer and they did not receive one. The refund is the whole of the remedy. No replacement reviewer is appointed, the version is not assessed a second time, and the author is not charged for what they did not properly receive. An author who wants a decision on the work revises, deposits a new version, and buys an assessment of that version on the ordinary terms.

7.3.9 · An appeal either succeeds or does not succeed.

7.3.10 · When a refund arises. An assessment fee is refunded on two grounds and no others: the report was not delivered, or an appeal against the certification decision succeeded.

7.3.11 · An assessment closed without a report is non-delivery, and is refunded in full. Where an assessment is closed before any report is delivered — including where it is closed following an integrity finding under the Retraction and Correction Policy — the author did not receive what they bought, and the assessment fee is refunded on the first ground in 7.3.10. This is not a separate ground and no new one is created; it is what non-delivery means. The author need not request it: XSci issues the refund itself, and no time limit applies. Refund mechanics are governed by the Refund Policy.

7.3.12 · Disputes about a correction or retraction, as distinct from a certification decision, are handled under the Retraction and Correction Policy.

8. Ethical standards

8.1 · What XSci expects. XSci expects everyone taking part in review — authors, reviewers, and staff — to observe applicable research ethics, publication ethics, legal requirements, and integrity standards.

8.2 · What the assessment record states. Which version was assessed, the target venue, the Standard version applied, the reviewer's declared interests, the number of reviewers and how many certified, the visibility of the report, and who paid for the assessment, so that a reader can see whether the purchaser was the person whose work was examined.

9. Contact

9.1 · For questions about the peer review process, contact editorial@xsci.com.

10. Policy updates

10.1 · Versioning. XSci may update this policy to reflect evolving services, workflows and best practice. Each version carries the date it was uploaded, in the header above.

10.2 · Changes apply going forward. Where a change to this policy would alter what a certification means, the change applies to assessments purchased on or after the date the new version is uploaded. Assessments already purchased are completed under the version in force when they were bought, and the assessment record names that version.

10.3 · Related documents. The XSci Certification Standard states the criteria in full. The Certification Procedure states how an assessment is conducted, how a report is checked, how an appeal is filed and decided, and how an integrity referral is handled. The Reviewer Agreement, AI and Data Usage Disclosure, Retraction and Correction Policy, Refund Policy, Privacy Policy and Terms of Service govern their own subject matter. Where any of those conflicts with this policy on what review decides and what certification means, this policy and the Certification Standard control.