Credibility, decomposed.
A single number can't capture trust. ScoRISK breaks credibility into five independent layers, scores each on its own evidence, and combines them with reasoning you can inspect.
Five independent layers of analysis
Each layer evaluates one dimension of credibility on its own evidence. Scores combine through a weighted aggregation model — never a single black-box number.
Content Verification
Core Truth LayerExtracts factual claims using NLP and cross-checks them against trusted news databases, fact-check APIs, and cross-article similarity detection. Surfaces verified, partially verified, unverified, and contradicted claims.
30% weightAuthor Credibility
Human Trust LayerEvaluates the person or entity behind the content — historical accuracy rate, domain expertise alignment, writing consistency, citation quality, and behavioral signals like clickbait usage or retractions.
20% weightSource Reliability
Origin Trust LayerAssesses the publication platform itself — domain reputation, editorial standards, ownership transparency, and history of misinformation involvement. Where something comes from matters as much as what it says.
20% weightLanguage & Manipulation
Psychological LayerDetects emotional intensity, certainty vs nuance, fear-based language, urgency triggers, and the ratio of logical structure to persuasion tactics. Misinformation is often not false — it is persuasive.
15% weightMedia Authenticity
Visual Truth LayerVerifies images and video through reverse image matching, deepfake probability estimation, metadata consistency analysis, and contextual mismatch detection. Visual content is often more misleading than text.
15% weightNever just a number.
The system always outputs a breakdown, a reasoning explanation, and a confidence level alongside any final score. Transparency is not optional — it's the product.
Each layer produces a normalized score from 0–100. These are combined via a weighted aggregation model that accounts for contradiction amplification, recency adjustment, virality risk, and cross-source consensus.
- Content VerificationClaims evidence quality30%
- Author CredibilityHistorical reliability20%
- Source ReliabilityPublication standards20%
- Language AnalysisManipulation patterns15%
- Media AuthenticityVisual verification15%
Smart, not static.
Beyond the five layers, ScoRISK continuously adjusts confidence based on what the information landscape is doing right now.
Contradiction Amplifier
When multiple trusted sources disagree, confidence is reduced and the overall score is adjusted downward to reflect genuine uncertainty.
Recency Adjustment
Breaking news is inherently uncertain. Confidence drops temporarily for newly published content, without unfairly penalizing the score itself.
Virality Risk Signal
Highly viral content with low verification gets flagged as 'fast-spreading unverified' — surfacing risk at the exact moment it matters most.
Consensus Factor
Independent agreement across multiple credible sources is a genuine signal of reliability. The system boosts scores when true consensus exists.
Everything that exists. Everything that's missing.
No single tool today combines fact-checking, source reputation, author credibility, language analysis, and media forensics into one explainable score. ScoRISK does.
| Capability | Fact-checkers | Browser extensions | Research tools | ScoRISK |
|---|---|---|---|---|
| Content fact-checking | Partial | Partial | Full | |
| Author credibility analysis | Full | |||
| Source reputation scoring | Partial | Full | ||
| Media forensics | Separate | Integrated | ||
| Language manipulation detection | Research only | Full | ||
| Unified intelligent scoring | Built for it | |||
| Real-time explainability | Core feature |
The Bloomberg Terminal
for information trust.
Not a fake news detector. A credibility intelligence system — the credit score for the information ecosystem.
- Daily-use information filter
- Browser intelligence layer
- Decision-support system
- Trust infrastructure
Ready to build the trust layer the internet needs?
Get early access to ScoRISK Credibility Intelligence and shape what information intelligence looks like.