Rules

Any researcher, in any discipline and any country. The submitting author must verify an ORCID iD or an institutional email address, and must do so before entries close. An entry whose submitting author is not verified at the close is not an entry for the purposes of these rules: it is not considered for expert review under rule 5, and rule 10 does not apply to it. The preprint, its permanent identifier, its priority date and its AI referee reports are unaffected, and the author may buy a Peer-Certified assessment in the ordinary way at any time.

The manuscript must not have been published in a journal or conference proceedings. Manuscripts posted as preprints elsewhere, or under review elsewhere, are eligible; checking for conflicts with another venue's policy is the author's responsibility.

All co-authors must have agreed to the entry and to the publication of the expert review report, which is what entry to the award turns on. Whether the AI referee reports are disclosed is a separate matter, decided under rule 4.2, and it does not affect eligibility either way.

At most three manuscripts per submitting author. A revised version of a manuscript already entered is not a further entry: it replaces the earlier version, and rule 5.5 says what follows.

The following may not enter: XSci staff and contractors, and their immediate collaborators; anyone taking part in selection under rule 5; and anyone previously removed from the competition, or from the XSci platform, for misconduct.

the manuscript is published as an XSci Preprint;

the AI referee reports are published on the preprint page. A report may still be taken down by paying the standard difference between the two prices, and the entry remains eligible: disclosure of AI review is independent of the competition, and nothing in it bears on entry or on the award;

if no target journal is named, the entry is assessed against the leading journals in its field, and the page shows this as the target.

entrants may purchase expert review in the ordinary way. Where such an entry is later selected for expert review under rule 5, it is not assessed a second time: the decision on the assessment already purchased stands, and that assessment occupies one of the places. Where the author makes that report public, the entry is eligible for the award on the same terms as any other.

Rolling places (up to 60). From the opening date, entries whose AI reports clearly place them as plausible candidates for the leading journals in their field are admitted in order of submission until the places are taken. Admission is automatic. An entry reached in that order whose submitting author is not yet verified is not admitted at that point, and is admitted when the verification is settled, in the order in which verifications complete. The place is neither reserved nor lost while the position stands open.

Remaining places. After the close, allocated to disciplines in proportion to entries received, then to the highest-ranked entries within each discipline.

Ranking. Uses the AI referee reports and the manuscript, and weighs novelty, rigour, clarity, and fit with the target the author has named. An entry that names an ambitious target and is judged a plausible fit ranks above an otherwise similar entry with a modest target. The method is published after the final results, not before.

Selection is not appealable. An AI referee report is not a verification of anything and certifies nothing.

Revising an entry. An author may deposit a revised version at any time, and revising is never held against an entry. A revision is not a further entry for the purposes of rule 3.4. What a revision does depends on where the entry has got to.

One expert per entry. Candidates are drawn from the authors the manuscript cites, and from authors who have published on the topic in the target journals. Experts are invited automatically. An expert is paid whatever the decision they reach. Payment turns on whether the assessment was carried out and reported as required, never on the outcome. The expert is anonymous unless they choose otherwise.

Conflicts of interest are decided under the Certification Procedure, clause 3.3, which applies here without modification. Four interests bar a candidate outright, under clause 3.3.1: a prior role in the work; a close personal relationship or dispute with an author; a financial or commercial interest in the outcome; a current supervisory relationship with an author. An expert who is an entrant in this competition, who is excluded under rule 3.5, or who is reviewing the same manuscript for another venue, is also barred. Everything else — including recent co-authorship, a shared affiliation, competing work and academic rivalry — is declared and recorded, not barred, under clause 3.3.3. Experts confirm their position on accepting. Where a barred interest emerges before a report is delivered, the entry is reassigned. Where one emerges afterwards, the certification is revoked under the Retraction and Correction Policy, clause 2.4.3, because the record is then materially false.

One decision, accept or reject, on the test in the Certification Standard. Accept means that, in the expert's judgement, the manuscript:

Where the author has named no target, the target is the leading journals of the manuscript's field, and the expert must identify and name them in the report. The bar is relative to the target named, and not absolute: a sound manuscript aimed higher than it can reach is not accepted, and the expert must say where the work would fit instead. Acceptance is not an independent verification of the results.

The form of the report is the one the Certification Procedure requires, under clause 5.2: a finding on each of the nine assessment domains the manuscript engages, a section on fit, and every reason stated specifically. There is no score, grade, rating or ranking, and "revise and I will judge then" is not an available decision.

One round, one expert, one version. The version assessed is the one fixed under rule 5.5(d) — the version standing when the expert accepted the assignment. The author may deposit revised versions afterwards; the decision rests on the version reviewed, and that version is not assessed again — except on a successful appeal under rule 9. Where the entry is certified, the mark attaches to the version assessed and carries forward to later versions under the Peer Review Policy, clause 6.9, which applies here without modification — authors are never asked to pay to certify a revision.

An accepted entry is an XSci Peer-Certified Preprint, whether or not the author keeps the expert report public.

The report is the expert's own judgement, and the expert is fully responsible for it. Any use of AI is governed by the Reviewer Agreement, Section 3, and is recorded on the assessment record. No machine certifies anything.

Integrity matters are referred, not decided. An expert who encounters plagiarism, undisclosed duplicate publication, or signs that images or data have been manipulated must refer the matter to XSci and stop, and must not write the allegation into the report. The entry is then handled under rule 11.

Published on the preprint page by default, with the decision.

During the buffer window (48 hours) after the report arrives, only the author can see it and may take it down at no charge. Nothing about it is then shown.

Taking the report down ends eligibility for the award. It does not affect certification: an accepted paper keeps its Peer-Certified mark.

Authors may reply to a published report on the page. A reply is commentary on the record and does not reopen the decision.

A judgement in a delivered report is never altered. Where an expert adds or corrects something afterwards, it appears as a dated addition, with the original retained alongside it.

Every accepted entry with its human expert report public receives the XSci Open Review Award 2026 and USD 500.

Paid to the submitting author by XSci LTD within 30 days of certification, by bank transfer or another method suited to the recipient's location. XSci bears transfer fees. Where payment cannot be made, the author's platform account is credited with USD 500 of services instead.

Recipients are responsible for any tax in their own country. XSci may ask for identification and payment details as its payment provider requires.

Four grounds only. An appeal against the expert's decision lies on the four grounds in the Peer Review Policy, clause 7.3.4, and no others:

There is no fee. An appeal is made by email within 14 days of the report.

Disagreement with the expert's judgement is not a ground, and that includes the judgement on fit and on the target journals.

How it is decided. By a person who took no part in the assessment. An appeal either succeeds or does not succeed.

The remedy is a fresh assessment, not a refund. Each of the four grounds is a finding that the assessment was not delivered as required. Where an entry was paid for, the Peer Review Policy answers that with a refund of the assessment fee, under clause 7.3.6; a competition entry carries no fee, so XSci supplies the assessment again, at its own cost, on the same version and with a different expert. The fresh decision stands in place of the first. This is an exception to the rule in the Peer Review Policy, clause 2.2.3, that a version is assessed once and once only, and it is an exception for this competition alone: it arises because there is no fee to refund, it operates only where an appeal has succeeded, and it changes nothing in the paid service.

What this is, and what it is not. It is the replacement of an assessment that was not delivered as required. It is not a second opinion on a valid report, and it is not available because the author disagrees with the decision. The first report is withdrawn from the record with a dated note that it was withdrawn on appeal, and the first expert's payment is not due, under the Reviewer Agreement, clause 8.4.

Where the successful ground is 9.1(b) and the entry had been certified, the certification is revoked under the Retraction and Correction Policy, clause 2.4.3. The revocation stays visible on the record, under clause 2.4.5. A fresh assessment under 9.5 may certify the same version again, and where it does, the new certification carries its own date and record.

An appeal against a revocation is not decided on these grounds. It is decided against the COPE, ICMJE and Crossref standards named in the Retraction and Correction Policy.

Grounds. XSci may remove an entry, withhold or reclaim an award, and note the reason on the preprint page where an entrant:

The author is heard first. Before anything is published, the submitting author is told the allegation and has 14 days to respond.

Certification is not withdrawn under this rule. Where an entry has been certified, the certification is revoked only under the Retraction and Correction Policy, clause 2.4.3, on the one ground stated there — that the record was materially false — and only in the manner that policy requires. A revoked certification is never deleted (clause 2.4.5): it stays publicly visible, marked as revoked, with the date and a reason.

While a matter is open, the entry simply shows as in progress, with no public reason given, and no notice identifies the manuscript as the subject of an enquiry. Where an integrity matter is raised during an assessment and made out, the assessment is closed and no report is delivered. Where it is not made out, the assessment resumes and concludes normally, and no record of the enquiry appears on the public record.

Experts. An expert who held an undisclosed barred interest, or whose report is not their own, is not paid and is removed from the pool, under the Reviewer Agreement, clauses 5.3.3 and 8.3. Nothing about an expert individually is published, then or at any time.

All versions and published reports remain on the record.