InkCop’s Citation Verification System is a multi-agent collaborative mechanism purpose-built for academic writing. It doesn’t just check citation formatting — it traces the actual existence of citation sources, eliminating hallucinated references at the root.

Why Dedicated Citation Verification?

When you use general-purpose LLMs (like ChatGPT, etc.) to assist with writing, the model may “invent” references that sound plausible but don’t actually exist — this is known as Citation Hallucination.

This is a critical risk in academic writing:

• Fake DOIs: Generated DOIs point to non-existent papers
• Wrong author/year combinations: Real authors with incorrect publication years or journals
• Mixed citation styles: APA, IEEE, Chicago styles mixed together or with missing fields
• Untraceable: No way to confirm the cited content aligns with the original paper’s argument

How InkCop Citation Verification Works

Layer 1: Structured Citation Parsing

When AI inserts a citation in the editor, InkCop automatically parses it into structured data:

Author list → Publication year → Journal/Conference → Title → DOI/URL

The system recognizes APA, IEEE, Chicago, Vancouver, and other major citation styles, extracts each field, and normalizes them for verification.

Layer 2: Multi-Source Real-Time Tracing

The Verification Agent queries multiple authoritative databases in parallel:

Layer 3: Semantic Consistency Check

Confirming a paper exists is not enough. InkCop goes further:

1. Abstract comparison: Retrieves the original abstract from the database and compares semantic similarity with the citation context
2. Argument alignment detection: Confirms the cited paper’s core argument matches your citation context, preventing “out-of-context” misquotes
3. Reverse source lookup: For papers already in your knowledge base, directly locates the relevant paragraphs in the original text

Layer 4: Standardized Format Output

After verification, the system automatically converts citations to your target format (APA 7th, IEEE, Chicago 17th, etc.), ensuring consistent style throughout the document.

Verification Status Labels

Each citation receives one of the following statuses:

• ✅ Verified: Paper exists, content consistent, format correct
• ⚠️ Needs Review: Paper exists but some fields are questionable (e.g., year, page numbers)
• ❌ Verification Failed: Cannot find a matching paper in databases (likely hallucination)
• 🔗 Local Match: Found the original paper in your local knowledge base

Integration with Your Local Knowledge Base

The citation verification system is deeply integrated with InkCop’s local RAG knowledge base:

• When citing papers already in your knowledge base, verification requires no internet connection — checked directly against your local database
• Verified citations automatically build a citation relationship graph between documents for future reference tracing
• All verification records are stored locally and never uploaded to any server

Comparison with Other Tools

Using Citation Verification

In the InkCop editor, you can:

1. Automatic trigger: Verification starts automatically in the background when AI inserts a citation
2. Manual trigger: Select any citation text and click the “Verify Citation” button
3. Batch scan: Use “Scan All Citations” in the document menu to check all citations at once

Tip: The Citation Verification Agent runs asynchronously in the background, so it never interrupts your writing flow. Results are displayed in a non-intrusive way once verification is complete.