Siftr editorial workspace

Find the strongest ideas in a research paper. Preserve the evidence, reasoning and context behind them.

Upload a new paper

Editorial queue

Submissions

Return to unfinished reviews, check progress, or reopen a completed editorial record.

Enter the administrator password to view saved submissions.

The Siftr workflow

From source paper to published insight

Each stage is designed to identify the paper’s most valuable insights without oversimplifying its reasoning. Siftr combines AI analysis, source mapping and human editorial judgement to make those insights clear, faithful and useful.

01

Submit and verify the paper

The paper is submitted with its PDF and DOI, then checked for source accuracy, publication details, research quality and relevance to Siftr’s domains, including health, economics, public policy, and systems and innovation. Siftr verifies that the file matches the DOI record and checks publication metadata, licence, integrity and scope.

02

Assess research suitability

The paper is checked for a clear research purpose, methods or analytical approach, findings, limitations and evidence locations. Unsuitable, out-of-scope or unverifiable papers do not proceed automatically.

03

Identify candidate Siftrs

AI applies Siftr’s editorial rules to identify the paper’s distinct, meaningful contributions. Each candidate must express one standalone insight, remain central to the authors’ reasoning, be supported by specific evidence, avoid duplicating another candidate, and preserve important qualifications and limitations.

04

Build the evidence record

Every candidate is linked to precise source evidence: section, page, table, figure or other stable locator. Siftr retains only the minimum source text required for verification.

05

Run an independent quality review

A separate evaluation checks whether the candidate is central to the paper, evidence-backed, distinct from other candidates, appropriately qualified and suitable for human review.

06

Human editorial decision

A reviewer verifies the evidence, interpretation, relevance and limitations, then approves, requests revision or rejects the candidate. AI cannot publish directly.

07

Publish with traceability

Only approved Siftrs enter the public collection. The published record keeps its source, review history, version and update status, while the uploaded PDF and temporary full text are deleted.

Editorial safeguards

Evidence before publication

Every published Siftr must have verified primary support that a reviewer can locate in the authorised source.

Human accountability

Reviewer identity, decision, notes and approval status are retained internally for editorial accountability.

Minimal source retention

PDFs and temporary full text are deleted after processing. Only metadata, evidence locations and limited verification material are retained.