How DigiTrust works

DigiTrust combines several credibility-signal groups into a transparent score. The result is designed for triage and media literacy, not as proof that a claim is objectively true or false.

1. Fact-check evidence

When a Google Fact Check Tools API key is configured, DigiTrust extracts a small number of factual-looking sentences and searches published ClaimReview results. Matching reviews are displayed as evidence.

2. Domain signal

The source hostname is evaluated with conservative heuristics for established publishers, public institutions, education namespaces, and generic publishing platforms. This is not a licensed reputation database.

3. Independent coverage

When NewsAPI is configured, DigiTrust searches for related reporting and excludes the analyzed source domain. No coverage is treated as a neutral signal rather than proof of falsehood.

4. Language signals

The built-in analyzer looks for clickbait, emotionally loaded words, absolute claims, vague attribution, excessive capitalization, and punctuation intensity. An optional LLM can provide richer language analysis when configured.

Scoring

The reference weighting is 35% fact checks, 25% domain signal, 20% independent coverage, and 20% language signals. Missing external components are explicitly marked unavailable and contribute a neutral baseline.