F-001
Figure similarity candidate
Figure 3 panels B and D require author review.
AI-assisted research integrity workflow
Pre-submission QC for research manuscripts, figures, and data
Research Integrity QC AI helps researchers and PIs identify potential manuscript, figure, statistical, and data-consistency issues before journal submission.
The platform provides AI-assisted screening and structured QC reports for human review. It flags potential issues; it does not determine fabrication, falsification, misconduct, or fraud.
Pilot use is limited to non-confidential, de-identified, or already-submitted materials. Do not submit identifiable patient data or unpublished source datasets through the public website.
Findings
12
High
1
Moderate
4
S-001
Statistical reporting mismatch
Methods describe SD; figure legend describes SEM.
D-001
Source-data alignment
Manuscript states n=8; source CSV shows n=7.
Downloadable QC report
Evidence-linked findings, calm severity labels, and recommended author review actions.
Problem
The hidden QC burden in modern research
Research groups generate manuscripts, figures, statistical analyses, and supporting datasets faster than PIs can manually verify every detail. Errors in figure assembly, labeling, statistics, sample-size reporting, and manuscript-data consistency may remain undetected until peer review, publication, public scrutiny, or retraction. Labs need a confidential pre-submission QC workflow before problems become reputationally damaging.
Checks
What the platform checks
The early workflow focuses on practical, reviewable signals that can help authors correct problems before submission.
Manuscript consistency
Checks for internal inconsistencies, missing methodological details, unsupported claims, and mismatch between abstract, results, figures, and conclusions.
Figure QC
Flags potential duplicated panels, labeling inconsistencies, missing scale bars, low-resolution images, and figure-legend mismatch.
Statistical reporting
Reviews p-value reporting, sample-size statements, statistical-test descriptions, error-bar definitions, and consistency between text, tables, and figures.
Data-manuscript alignment
Helps compare reported findings against uploaded tables or structured supporting data when provided.
QC report generation
Produces a structured report with issue type, severity, confidence level, location, and recommended author review steps.
Workflow
How it works
Public forms collect only pilot-interest information. Any file review should happen later through an approved pilot workflow.
Request pilot access
Researcher submits a pilot request through the website.
Confirm material suitability
Early pilots use non-confidential, de-identified, or already-submitted material only.
Upload through secure workflow
Upload should only occur after pilot confirmation, not through the public contact form.
AI-assisted QC review
Manuscript, figure, statistics, and data-consistency modules flag potential issues.
Receive QC report
The researcher receives a structured report for human review before submission.
Sample report
Built around concrete report value
The sample package includes seeded issues so potential design partners can evaluate the report format without sharing unpublished or sensitive data.
- Manuscript claim drift and missing verification statements.
- Figure panel similarity candidates for human review.
- SD/SEM and sample-size consistency checks.
- Downloadable report template for concierge alpha pilots.
Trust and safety
Built for confidential scientific review
The product direction is quality improvement, not accusation. The public website is intentionally limited to discovery and pilot screening.
Human review first
- The platform is not a misconduct investigation tool.
- It does not determine fabrication, falsification, or fraud.
- It flags potential issues that require human review.
Public form limits
- Public forms must not collect confidential manuscripts, patient identifiers, or unpublished source datasets.
- Early pilots should use de-identified, non-sensitive, or already-submitted materials.
Future secure upload
- Authenticated upload workflows should include encryption, access control, audit logs, retention limits, and deletion options.
- User-uploaded materials should not be used for model training without explicit written consent.
Founder
Led by biomedical research experience
Research Integrity QC AI is led by Xi-Long Zheng at the University of Calgary and is being developed from the perspective of biomedical research quality improvement.
The platform is being developed as a research-quality improvement workflow: a way to help authors, labs, and research offices find correctable issues before submission, not a tool for public accusation or misconduct determination.
Upload policy
File upload is available only after pilot confirmation
The public website does not accept manuscript files, raw datasets, patient information, reviewer materials, or confidential research records. Future upload workflows should be limited to approved pilots and protected by security controls.
- Authenticated user accounts and project-level access control.
- Encrypted storage and transfer for approved pilot materials.
- Clear consent, file-type restrictions, and malware scanning.
- Audit logging, deletion options, and retention limits.
- No model training on user materials without written consent.
Pilot access
Request pilot access
Use this early form to start a discovery conversation. Do not send confidential manuscripts, identifiable patient data, or unpublished source datasets through this page.
Positioned as quality improvement, not accusation
This workflow flags potential issues for human review. It does not determine misconduct, fabrication, falsification, image manipulation, plagiarism, or fraud.
Contact Xi-Long Zheng