Paper Mill Research Integrity Screener

Manual manuscript screening for research integrity is slow and prone to oversight. This skill automates the detection of common red flags like tortured phrases, citation mismatches, and statistical discrepancies to streamline the editorial triage process. It provides a technical, evidence-based report that helps human reviewers focus on high-risk areas without jumping to conclusions about intent.

What it does

• Identify tortured phrases by flagging awkward technical paraphrases and unusual word substitutions that suggest machine translation or obfuscation.
• Audit citation integrity by cross-referencing in-text citations with the reference list and checking for irrelevant or padded metadata.
• Verify numerical consistency by reconciling sample sizes, p-values, and percentages across the abstract, tables, and body text.
• Detect template indicators by spotting boilerplate ethics statements, mismatched methods, and structural inconsistencies common in industrial paper mills.
• Assess evidence severity by categorizing findings into critical review triggers, major anomalies, or minor documentation gaps.

How it works

1. Scope definition: Upload the manuscript and available supplements to establish the screening boundaries and missing materials.
2. Automated screening: The skill scans the text for terminology anomalies, statistical errors, and structural patterns.
3. Evidence logging: Every observation is recorded with its exact location, a minimal excerpt, and a confidence rating.
4. Triage reporting: A final report is generated including an executive summary, finding logs, and a human reviewer handoff checklist.

Frameworks & tools

This tool relies on linguistic analysis and internal consistency checks. It is designed to work with standard academic manuscript formats including PDF, Word, and LaTeX.

Why this beats prompting it yourself

Generic prompts often produce speculative "AI-generated" verdicts or miss subtle numerical discrepancies. This skill enforces a strict evidence-based workflow that requires location-based proof, considers alternative explanations, and prevents the agent from making definitive claims of fraud.

Use cases

• Pre-peer review screening to identify low-quality or suspicious manuscripts before assigning reviewers.
• Research integrity audits for institutional repositories or journal archives.
• Statistical sanity checking to catch accidental reporting errors in complex data tables.
• Paper mill detection by identifying repeated structural boilerplate across multiple submissions.

Known limitations

Does not prove author intent or misconduct. Requires manual verification of external citations if DOIs are not provided. Cannot definitively identify AI authorship from prose style alone.