Skip to main content

Verification & Trust

Argus provides multi-layered verification to help you assess content reliability.

Confidence Scores​

Every article gets a confidence score (0-100) based on:

  • Source reliability - Historical accuracy of the publication
  • Credibility indicators - Citations, specific data, expert quotes
  • Claim verification - Cross-referenced with other sources
  • Bias analysis - Emotional language, sensationalism, political lean

Score Levels​

LevelScoreMeaning
High80-100Well-sourced, factually accurate, neutral tone
Medium60-79Moderately reliable, consider cross-referencing
Low40-59Exercise caution, verify key claims
Very Low0-39Significant concerns, verify with trusted sources

Ground Truth Sources​

Wire services are treated as ground truth due to their journalistic standards:

  • Associated Press (AP News)
  • Reuters
  • AFP (Agence France-Presse)

Claims corroborated by wire services get an automatic confidence boost.

Claim Extraction​

Argus extracts factual claims from articles using AI:

# Extract and verify claims for an article
curl -X POST "https://argus.vitalpoint.ai/api/verification/verify-claims/{contentId}"

Each claim is assessed for:

  • Verifiability - Can it be fact-checked?
  • Status - verified, partially_verified, unverified, contradicted
  • Corroboration - Which other sources support or contradict it?

Cross-Reference Verification​

Claims are automatically compared against your article database:

# Cross-reference all claims for an article
curl -X POST "https://argus.vitalpoint.ai/api/verification/cross-reference/content/{contentId}"

A claim is marked verified when:

  • Found in 3+ independent sources, OR
  • Corroborated by a wire service (ground truth)

Bias Detection​

AI-powered analysis of political lean and journalistic quality:

# Analyze article bias
curl -X POST "https://argus.vitalpoint.ai/api/verification/bias/{contentId}"

Returns:

  • Political bias: far-left to far-right spectrum
  • Emotional language: none/low/medium/high
  • Sensationalism: clickbait detection
  • Specific indicators: loaded language, unsupported claims, ad hominem attacks

Verification Trail​

See exactly why an article got its confidence score:

# Get full verification trail
curl "https://argus.vitalpoint.ai/api/verification/trail/{contentId}"

Returns a step-by-step breakdown:

  • Source reliability contribution
  • Claim verification results
  • Cross-reference matches
  • Bias indicators
  • Overall recommendation

Deep Verification​

Run the full verification pipeline in one call:

# Full verification: claims + cross-reference + bias + trail
curl -X POST "https://argus.vitalpoint.ai/api/verification/deep/{contentId}"

This is expensive (multiple LLM calls) but provides comprehensive analysis.

Batch Operations​

Batch Claim Extraction​

curl -X POST "https://argus.vitalpoint.ai/api/verification/claims/extract-recent?limit=10"

Batch Cross-Reference​

curl -X POST "https://argus.vitalpoint.ai/api/verification/cross-reference/batch?limit=20"

Batch Bias Analysis​

curl -X POST "https://argus.vitalpoint.ai/api/verification/bias/batch?limit=20"

Statistics​

# Overall verification stats
curl "https://argus.vitalpoint.ai/api/verification/stats/overview"

# Cross-reference stats
curl "https://argus.vitalpoint.ai/api/verification/cross-reference/stats"

# Source bias summary
curl "https://argus.vitalpoint.ai/api/verification/bias/source/{sourceId}"

Best Practices​

  1. Trust but verify - High-confidence scores are a good signal, but always check important claims
  2. Wire services first - Prioritize AP, Reuters, AFP for breaking news
  3. Check the trail - Use /trail to understand why a score was assigned
  4. Bias awareness - Use bias analysis to understand perspective, not to dismiss content
  5. Cross-reference important claims - Run deep verification on high-stakes content