Part of the CheckRef platform · Built on Crossref, OpenAlex & ORCID

See the evidence before peer review

SciReview gives editors a structured, evidence-based pre-review of every manuscript — structure, integrity, similarity, ethics and more — so editorial decisions are faster, fairer and fully traceable.

It never accepts or rejects. Every finding shows its evidence. The editor decides.

scireview.checkref.org/scireview/dashboard
EditorBrief · MS-2026-0488 ● Manual review recommended
Duplicate publication · deterministic + metadata
Possible related publication from 2023 — different cohort. The authors' own preprint matched as expected and was set aside.
Title similarity71%
Abstract similarity58%
Shared authors1 of 4
Self-match · medRxivset aside
Confidence0.74
Recommended: open the 2023 article and confirm it reports a different study before assigning reviewers.
SciReview never decides. It shows you why — and hands you the call.
The gap

Desk screening is the editor's heaviest task — with the least support

Before a manuscript ever reaches reviewers, an editor must judge whether it's complete, original, ethical and ready. Today that happens by hand, under time pressure, with tools that either say nothing or say too much.

Problems surface late

Missing ethics statements, prior publication, or structural gaps are often caught only after reviewers have spent their time — or after publication.

Screening is inconsistent

Two similar manuscripts can get very different first reads depending on who is on desk that day and how busy the office is.

Black-box AI erodes trust

"AI detectors" and one-click verdicts make claims editors can't verify — and that authors can't fairly answer. Trust needs evidence, not a score.

What SciReview does

Evidence first, always — never a verdict

SciReview runs a manuscript through independent review engines and assembles one editorial brief. Every line traces back to something you can check.

01

Every finding shows its evidence

No bare "high similarity detected." Each finding lists the matching evidence, a confidence estimate, and a recommended action you can act on.

02

Multiple models, reconciled honestly

When two AI reviewers disagree, SciReview surfaces both positions and recommends a human read — it never quietly picks a winner.

03

Eleven review engines, one brief

Structure, integrity, similarity, duplicate publication, ethics, reporting guidelines and more — deterministic checks first, AI to explain them.

04

The editor stays in control

SciReview recommends; it never accepts or rejects. The decision is yours, recorded with the brief and its full provenance.

How it works

Five steps from submission to an editorial brief

1

Ingest

Upload the manuscript (DOCX, PDF, TXT). SciReview parses sections, references and metadata.

2

Review

Deterministic engines run first; interpretive engines add AI judgement on top, only where it helps.

3

Reconcile

Where models are used, their findings are compared; disagreement is surfaced, never hidden.

4

Assemble

Findings become one EditorBrief — evidence, confidence and recommended actions, with provenance.

5

Decide

You review the evidence and make the call. Send to review, request revisions, or return — your decision, logged.

Review engines

Independent engines, each answering one question

Deterministic engines are reproducible and never call a model. Interpretive engines use AI to explain the evidence. One engine is advisory only — and stays out of the summary by design.

Deterministic

Structure

Is this a complete manuscript? Detects every expected section and flags what's missing.

Interpretive

Integrity

Surfaces possible integrity concerns for a human to confirm — observations, never accusations.

Deterministic

Similarity

Section-level overlap and source breakdown, with optional Turnitin / iThenticate integration.

Deterministic

Duplicate publication

Searches Crossref and OpenAlex — and sets aside the authors' own preprints, so you see real matches only.

Interpretive

Scientific writing

Publication-readiness of the language — framed as editorial support, never a quality gate.

Advisory

AI-assisted writing

Surface observations only. No score, never an accusation, and excluded from the summary.

Deterministic

Reporting guideline

Detects the study type and checks CONSORT, PRISMA, STROBE and other checklists for gaps.

Deterministic

Ethics & compliance

Ethics approval, consent, registration, conflicts, funding and data availability — present or missing.

Future

Statistical & revision

Statistical review and version-to-reviewer comparison are on the roadmap, flagged honestly as they mature.

Who it's for

Built for everyone who decides what enters peer review

📰

Editors

Start each manuscript with a structured brief instead of a blank page. Spend your judgement where it matters, not on triage.

⚖️

Editors-in-chief

Set the review profile, choose which models your journal trusts, and keep screening consistent across the whole editorial team.

🏛️

Publishers

Give every title the same evidence-based pre-review, with full provenance and an audit trail for editorial accountability.

The CheckRef platform

Five products, one scholarly record

From a clean reference list to a confident editorial decision — SciReview is where the manuscript is read before it ever reaches a reviewer.

Questions

What editors ask first

No — never, by design. SciReview assembles evidence, confidence and recommendations. Every editorial decision belongs to the editor and is recorded alongside the brief.

No. AI text detection is unreliable and unfairly flags non-native English writing. SciReview's AI-assisted-writing engine produces no score, is clearly advisory, never feeds the summary or recommendation, and must never be used to accuse an author.

Yes. Manuscripts are confidential third-party work. SciReview offers a local-only deployment with no external model calls, and where commercial models are used, only under zero-retention, no-training terms. Every external transmission is logged.

That's your choice. An editor-in-chief enables the providers the journal trusts — for example ChatGPT, Gemini or Claude — or runs none at all. Model names are configuration, never hard-wired.

No. SciReview is a pre-review that helps you decide what deserves to enter peer review. It augments editorial judgement; it does not replace editors or reviewers.

Fairness is a built-in principle. Language quality is framed as editorial support and is never used as a gate, and AI-writing observations are never treated as evidence of wrongdoing.

Give your editors evidence,
not guesswork.

Register your journal and run your first manuscript through SciReview. See the brief before you assign a single reviewer.