Certification Standard

This is the whole of what an XSci Peer-Certified Preprint claims, and how that claim is reached. It is published before any assessment is sold, and it is dated, so that anyone who sees the mark can look up what it meant on the day it was issued. Terms in bold on first use are defined in the appendix.

1. The claim

1.1 · What the mark says. An XSci Peer-Certified Preprint says this, and only this:

A qualified independent expert read this version of the manuscript in depth and judged that, after revision of the kind a journal would normally ask for, it has the potential to be published in the target journals the authors named — or, where they named none, in the top journals of its field — and that it contains no major error.

1.2 · What every certification shows. Every certification shows the date, the version it was issued against, the target journals it was measured against, and the version of this Standard in force.

1.3 · What it does not say.

• It is not an acceptance. No journal is bound by it. Whether a paper is published also depends on timing, competition for space and editorial priorities, none of which anyone can predict.
• It does not say the findings are true, or that they would come out the same way if someone gathered new data. That is not something a reviewer can establish by reading, and we do not claim it.
• It is one expert's honest opinion, formed on the manuscript in front of them. Another expert might have judged differently. It is not a guarantee.
• It is not a score. There is no grade, no rating, no ranking and no number. A manuscript is a certified preprint, or it is a preprint.

2. The two questions a reviewer answers

2.1 · Both must be answered yes. Certification requires yes to both. Either answer alone is not enough.

2.2 · First — is there a major error? Does what was done support what is claimed? Nearly every manuscript needs work, so the real question is whether what it needs is revision or reconstruction. Anything the authors could reasonably be asked to fix is ordinary feedback and does not stand in the way of certification. A major error does: the design cannot support the claims made, the methods cannot be evaluated at all, the numbers cannot be reconciled with the study described, or the central claims cannot be brought within the evidence by revising.

2.3 · Second — could it reach the target? Assuming that revision were done well, would this manuscript have a genuine prospect of publication in the journals the authors named? This is a judgement about the level of the contribution, and it is what an expert can offer that a machine cannot.

2.4 · The two are independent. A manuscript can be entirely sound and still not be certified because it was aimed higher than it reaches. That is a normal outcome, not a criticism of the work, and the reviewer says where it would fit instead.

3. The target journals

3.1 · Naming the target. Authors may name up to three journals when they buy an assessment. Naming is optional but strongly recommended.

3.2 · A list is a tier. Potential for publication in any one of the journals named is enough.

3.3 · Leaving it blank is a real choice, not a gap. Where no journal is named, the target is the top journals in the manuscript's field. The reviewer decides what those are, applying the same judgement they would to a named journal, and must say in the report which journals they had in mind. Many authors genuinely do not know where their work belongs, and for them a reviewer's view on placement is among the most useful things in the report. There is always a target on the record — a blank simply resolves to a broad one.

3.4 · The target is a gate, not advice. A manuscript named for a journal it does not reach is not certified, however sound it is. This is deliberate. Because the target is both the bar and part of the public record, aiming too high produces a refusal and aiming too low produces a mark that says exactly what it was measured against — so the sensible thing is to name the journals you honestly believe in.

4. What the reviewer examines

4.1 · A finding and a sentence on each area. The reviewer records a finding and at least one written sentence on each of these that the manuscript engages. The sentences are the report; they are what the authors can act on.

4.2 · These nine work in any field. Where one names something a manuscript does not have, it is marked not applicable — which is never a mark against the work. For theoretical, mathematical and qualitative work the same nine are read in the terms of that work: assumptions rather than sampling, derivations rather than procedures, whether a proof is valid and no wider than what it supports, whether an interpretive method is applied consistently and its basis shown.

4.3 · Verifiability means how far the reviewer got, and nothing more. They record the highest of these that holds: that the materials are declared; that they are reachable and as described; or that the reported results follow from them. That last step is reached differently in different work — re-running code where any is provided, recomputing figures from data, following a derivation from its assumptions, tracing quoted material back to its source, checking values against instrument or archival records. No kind of work is privileged: a manuscript with no data and no code is not less verifiable for it. Where materials were not provided, the reviewer says so, and that absence is recorded as an absence rather than treated as a fault. Whether findings would hold up on new data is never assessed and never implied.

4.4 · Alongside the nine, the reviewer must give their judgement on fit — the target journals, whether the manuscript has the potential to reach any one of them, why, and where it would fit instead. That last part is required whenever the answer is no.

5. Six things that prevent certification

5.1 · Four ordinary review findings. Four are ordinary review findings, and only prevent certification if they are still there after the authors have had their say:

• Required ethics, consent or regulatory approvals absent, or inconsistent with the work described.
• Methods so incompletely described that the analysis cannot be evaluated.
• Reported quantities irreconcilable with the stated design or sample, and not corrected.
• Central claims unsupported by the evidence presented, and not narrowed.

5.2 · The judgement is the reviewer's, and it is field-specific. What satisfies each of these differs by field, and that judgement is the reviewer's: what counts as an approval, or as an adequate description of method, is not the same in clinical medicine as in history or mathematics.

5.3 · Two are different in kind — undisclosed competing interests material to the findings, and plagiarism, undisclosed duplicate publication, or signs of image or data manipulation that the authors do not explain. A reviewer who sees one of these refers it to XSci and stops. They do not decide it, and they must not write it into a report that may be published, because a report is not the place to make an allegation. While such a referral is open, the assessment shows as in progress with no public reason: no one is flagged before a finding has been made. These matters are handled separately, with protections for everyone involved.

6. How an assessment runs

6.1 · The six steps.

1. The purchaser buys an assessment of one version and names the target journals, or leaves them to the default.
2. One reviewer is matched to the work and applies this Standard.
3. The reviewer records the decision, on the version presented to them.
4. The report is checked by XSci for completeness — never for its conclusion — and released to the authors with the decision, with any major error identified.
5. The authors have one response. It is recorded on the record beside the report as commentary; it does not change the decision or the report.
6. A revised version is a new deposit and a new assessment, bought on the ordinary terms.

6.2 · One version, one paid review, one reviewer. A second opinion on the same version cannot be bought at any price.

6.3 · No one at XSci may ask a reviewer to change a decision, and no one has the authority to.

6.4 · If certification is refused, there are two routes, and no third.

• Appeal. There is no fee. An appeal can be made on four grounds: the Certification Procedure — the published document setting out how an assessment is carried out, step by step — was not followed; the reviewer had an interest that should have barred them; the report contains an error of fact about the manuscript; or the report falls below the standard of detail and quality required — the areas in clause 4 not properly addressed, sentences that say nothing specific about this manuscript, or a placement judgement given without reasoning. Disagreeing with the conclusion is not a ground. That a reviewer weighed the evidence differently, or placed the work at a different level than the authors would, cannot be appealed. Where an appeal succeeds, the remedy is a refund of the assessment fee — see 6.5.
• Revise and come back. Post a new version and buy another round at the same fee. This is what the service is for: improving a manuscript until it is of publishable standard is the outcome it exists to produce, and coming back with revised work is not treated as anything other than that.

6.5 · What happens where an appeal succeeds. The remedy is one remedy, whatever the ground: the assessment fee is refunded in full. There is no further review, no replacement reviewer and no second judgement. The report stands as written, the record stands as it is, and the money is returned.

6.6 · An appeal is therefore not a route to certification. An author whose appeal succeeds is an author who paid for something XSci did not deliver properly and has been repaid — not an author who has been certified. The decision does not change on appeal. A refunded author is free to buy another round on a new version, on exactly the terms anyone else would get.

6.7 · Two words, used consistently. An appeal succeeds or it does not succeed — in this Standard, in the Guides, and in anything else XSci writes. We avoid the traditional appeal vocabulary here deliberately, because one of its words is used inside the platform to mean the opposite of what a reader would take it to mean.

6.8 · One consequence follows. An appeal was the only way a second reviewer could ever have come to read one version, and it is not that any more. The rule in 6.2 therefore holds in every case: a version is read once, by one reviewer, and no route reopens it. This is true whatever the ground and whatever the outcome — including where an appeal succeeds and the certification is revoked under clause 11. The remedy is money, never a second reading.

6.9 · A note on what may change. A re-assessment of the same version by a different reviewer, where an appeal has succeeded, is a remedy XSci may be able to offer in future. It is not available now, and nothing here should be read as promising it. If it is ever introduced it will be by a published change to this Standard, applying only to assessments bought after that change.

7. What is on the record, and who decides what is public

7.1 · What the record holds. Every completed assessment has a record. It holds: the version and its identifier; the date; the target journals; the decision and, where certification was refused, the reasons; the reviewer's finding and sentence for each area examined; their judgement and recommendation on fit; whether the authors responded; who paid; the reviewer's competing-interest declaration; whether the reviewer chose to see any AI review or earlier reports; and the reviewer's own use of AI in preparing the report.

7.2 · Publication of an assessment is the authors' decision — for AI review and human review alike.

• A certification is public, because displaying the mark is the authors' own choice. The record behind the mark is published with it.
• A refusal leaves no public trace unless the authors choose to publish it. XSci does not publish it and does not say whether a manuscript has been assessed at all.
• Authors who do publish a refusal do so at no cost: it is clearly marked as a refusal, carries no mark, includes the reviewer's recommendation of where the work would fit, and may be published with the authors' own reply beside it.

8. Reviewers

8.1 · Where they come from. For each manuscript, XSci invites experts from three groups: authors of work the manuscript cites, close collaborators of those authors, and researchers who have published on similar questions in the journals the authors are aiming at. That last group matters most for judging fit, and matching prefers them.

8.2 · Four things require a reviewer to decline outright, and disclosing one does not cure it: a prior role in this work (having advised on it, reviewed it for a journal, or taken part in producing or correcting it); a close personal relationship or a dispute with an author; a financial or commercial interest in the outcome or the subject matter; or a current supervisory relationship with an author in either direction. One narrow exception: the reviewer of a certified version may assess a later version of the same manuscript.

8.3 · Everything else is declared, not barred — recent co-authorship, shared affiliation, competing work in the same area, a rivalry, a strong prior public position. The reviewer declares it, XSci decides whether to proceed, and the declaration goes on the record. This follows the practice set out by the ICMJE and by major publishers. A reviewer who discovers a conflict partway through must stop and disclose it, and bears no consequence for doing so.

8.4 · Reviewers take no credit on the manuscript. Assessing is paid work, and never authorship, contributorship or acknowledgement.

8.5 · Reviewers are paid the same whatever they decide. No part of anyone's fee or compensation depends on the outcome. There is no contact between reviewer and authors in either direction; everything goes through the report.

9. Fees

9.1 · Two amounts, kept apart. The assessment fee is what an author pays XSci for an assessment. The reviewer payment is what XSci pays the reviewer who carries it out. They are separate amounts and neither is calculated from the other; this clause 9 concerns the assessment fee. A reviewer's payment never varies with the decision they reached; whether it is payable at all turns on whether the assessment was carried out and reported as required, and is governed by the Reviewer Agreement. An author's refund is owed by XSci in full and does not depend on it.

9.2 · The assessment fee buys the assessment. It never buys the certification. A refusal is a complete, delivered piece of work, not a failure of the service, and it is not refunded for being one. Refunds exist on two grounds and no others: non-delivery, and a successful appeal (clause 6).

9.3 · What the fee never depends on. It does not vary with the decision, it does not vary with the round number, and nothing can be added to it. A revision round is neither cheaper nor dearer than a first round: there is no discounted continuation rate, no premium option, and no extra a purchaser can buy. Because a second opinion on one version is not for sale at any price, the fee cannot be spent on improving the odds.

9.4 · What the fee does depend on is the authors' disclosure choice. There are two prices for a round of human review, and the authors choose between them at purchase by electing whether the report will be disclosed. The disclosed price is the lower of the two: the reduction is a discount given for publication, not a surcharge levied for privacy, and it is the only thing about the fee that varies. Amounts are published separately on the pricing page.

9.5 · Four rules go with that choice:

• A report bought at the discounted price is published automatically, 48 hours after the report is released to the authors — not 48 hours after purchase. The clock starts when the authors receive the report, so they always see it before anyone else does. They do not need to do anything, and there is no step at which they confirm it. This is what the discount is given for, and it is stated wherever the discounted price is offered.
• Withdrawing disclosure costs the difference between the two prices, because the discount was conditional on the disclosure and the condition has ended.
• That difference is charged once only. Authors who withdraw disclosure, restore it, and withdraw it again pay nothing further.
• It is calculated at the prices in force on the date of purchase, never at a later price. A change to the price list does not reach back into a round already bought.

9.6 · Moving the other way is free and is never refunded. Authors who bought at the higher, undisclosed price and later publish the report pay nothing and are repaid nothing. Opening a report always costs nothing; closing one costs the difference, at most once.

9.7 · Who paid is on the record. The record names who paid, so that a reader can see whether the person who bought the assessment was the person whose work it examined.

10. Versions, and how long a certification lasts

10.1 · A certification is issued against the version that was read, and the record names that version permanently.

10.2 · The mark carries forward. When the authors post a later version, the certification stands. It is not revoked, and authors are never asked to pay to re-certify a revision — revising is what authors are asked to do, and a mark that punished them for it would work against its own purpose.

10.3 · A substantially changed version says so. On posting a new version, the authors state whether it is a minor or a substantial revision of the certified one; where it is substantial, the page says so beside the mark. For example: "Certified at version 2 on 3 March 2027. Version 4 is the current version and is a substantial revision of the version certified." No claim is ever made that a later version was read.

10.4 · No expiry. Assessment of a later version can be bought at any time, at the same fee as any other. There is no expiry: the date belongs to the assessment, not to the certificate.

11. Correcting and revoking

11.1 · The word used here is revoked. A certification that is taken off the record is revoked, in this Standard and in every XSci policy. The word withdrawn is used on XSci for two unrelated things — an author withdrawing the disclosure of a report (clause 9), and an author withdrawing a deposit — and it is not used for a certification.

11.2 · A certification is revoked only on a finding that the record was materially false — that the work was not as the record described it, or that the record did not disclose an interest that barred the reviewer who made it. Never because a reviewer changed their mind, because a later study disagreed, or because the manuscript became controversial. A revoked certification is never deleted: the record stays, marked as revoked, with the date and a reason.

11.3 · Where the falsity is ours and not the authors'. An undisclosed barring interest is a failure of XSci's process, not of the authors' honesty. So two things happen together: the certification is revoked, and the authors are refunded in full, because what they bought was an assessment by an eligible reviewer and that is not what they received. The refund is the whole of the remedy — there is no replacement reviewer and no second reading of the version — and nothing about the reviewer individually is published. A revocation on this ground carries no implication whatever about the authors or the work.

11.4 · How a revocation is decided. Against 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. An author may appeal a revocation, without a fee; that appeal is decided against those same standards, and the remedy is that the revocation is set aside and the certification restored, or that it stands. No refund arises from a revocation appeal, because none is owed — where a refund was due on the revocation itself, it has already been paid. The detail is in the Retraction and Correction Policy.

11.5 · Factual errors on a record are corrected, with the original kept visible alongside. Judgements are never corrected. A reviewer's finding or decision stands as it was; a reviewer who wants to add something says it as a dated addition.

12. How the mark may be shown

12.1 · Two rules bind everywhere, on XSci and off it. The mark must never appear on an uncertified or revoked version. And it must never appear with a number, a tier, a percentage, a star, a band or comparative language, and must never be described as a rating, a score, a grade or a level of assurance. These two are absolute: they are what stops a certification being turned into something it is not, and they hold wherever the mark is shown.

12.2 · Two further rules govern how the mark is captioned: it should show the version it was issued against and the date, and it should link to the record. On XSci these two are binding. Off XSci they are what we ask of anyone displaying the mark, and what we recommend — we state this plainly rather than assert a control over third-party sites that we could not exercise without litigation.

12.3 · Nothing else XSci produces may be called certification — an AI review report is not certification, and no machine-generated output of any kind is.

12.4 · There is one mark. A manuscript is accepted as an XSci Peer-Certified Preprint, or it is a preprint. Any new mark would require a published change to this Standard, approved before anything was sold against it.

13. Changes to this Standard

13.1 · How changes are made. Versions are numbered and dated, with a note of what changed and why. Every version keeps its own permanent address, and the text of older versions stays available, because a certification issued under one version means what that version said. Changes apply only going forward. No existing certification is reinterpreted, and substantive changes are published before they are applied.

Appendix — Terms used here

Assessment. One purchased application of this Standard to one version of a manuscript, ending in a decision. It is the thing that is bought. It is not the same as a certification, which is one of the two things an assessment can produce.

Certification. The judgement in clause 1, and the public mark and record that follow from it.

Major error. A defect which, while it stands, means the manuscript's central claims are wrong, unsupported or impossible to evaluate — as distinct from something the authors could reasonably be asked to put right. The first four items in clause 5 are major errors by definition. Anything ordinary revision would address is not one.

Preprint. A manuscript the authors have posted publicly, with its own permanent identifier, before or without publication in a journal. Posting on XSci is never conditional on buying anything, and no one judges the work before it appears.

Record. The permanent statement of a completed assessment, listed in clause 7. Publication of it is the authors' decision; where a mark is displayed, the record behind it is published too.

Report. What the reviewer writes: a finding and at least one sentence on each area examined, the judgement on fit and where the work would fit, the decision, and the reasons for any refusal.

Target journals. What the authors are aiming at, recorded when they buy the assessment — up to three, or a blank that resolves to the top journals of the field. See clause 3.

Underlying materials. Whatever a manuscript's claims rest on, and which another competent person would need in order to check them. What this is depends entirely on the field: measurements, data, code or analysis scripts, transcripts, field notes, source texts, images, instrument or archival records, derivations and proofs, protocols, specimens, or the field's equivalent. Nothing in this Standard requires materials of any particular kind, and no kind is treated as better evidence than another.

Version. One posting of a manuscript, with its own permanent identifier, together with anything posted alongside it. Each revision is a new version.

Authors: see the Guide for Authors for how to buy an assessment, what happens next, and how to respond to a report. Reviewers: see the Guide for Reviewers and the Reviewer Questionnaire, which is the form this Standard is applied through.