News Intelligence System — Multi-AI consensus scoring for news credibility.
Five AI models — ChatGPT, Claude, Gemini, Grok, Perplexity — evaluate every article independently. Parallel processing. Peer review pairing. Perplexity gets final say via live web search. A 6th AI synthesizes into actionable bullets. ~$0.05–0.10 per article. 15–30 seconds. Production-ready.
The Scoring Process
- Health Check — Verify all 5 AI APIs are live
- Individual Scoring — Each AI scores 0–100 on accuracy, credibility, relevance, bias, quality
- Peer Pairing — AIs compare scores, Perplexity always participates for web verification
- Final Verdict — Perplexity cross-checks claims against live sources
- Consolidation — 6th AI call synthesizes into clear actionable bullet points
Why 5 Models?
Each AI has different training data, biases, and strengths. When 4 of 5 models score an article 85+, the confidence isn’t additive — it’s multiplicative. Disagreement is also valuable: sharp divergence signals the story is nuanced or ahead of verified data.
Key Metrics
- 5 AI models per article
- ~$0.08 average cost per article
- ~20 seconds processing time
- 5 scoring dimensions
- Google Sheets integration built in
- GUI available with progress tracking
Open Source (MIT): github.com/jjshay/news-intelligence · Back to Applied AI