Source it. Prove it.
Know who spread it.

// ENGINE 01 — PROVENANCE

Trace it to the source.

Every post has an earliest known instance. The Provenance Engine fingerprints content with SimHash for text and pHash for images, then queries Archive.org, TinEye, and Bing Visual Search for the first match. If the post is a repackaged 2019 image, you'll see the date stamp.

Archive.org · TinEye · Bing Visual Search · pHash · EXIF

// ENGINE 02 — SOCIAL GRAPH

Audit the network.

Bot detection and coordinated-amplification analysis on the account that posted. Follower-ratio anomalies, network topology, and the Misinformation Propagation Index — a persistent score tracking each account's historical role in disinfo campaigns.

Neo4j Graph DB · Botometer · MPI

// ENGINE 03 — REPUTATION

Score the source.

Composite credibility scoring on the account and the linked domain — bot signals, NewsGuard ratings, domain WHOIS age, Safe Browsing flags. Returns a trust band from VERIFIED to KNOWN BAD ACTOR with the evidence chain.

Botometer · NewsGuard · RDAP · Safe Browsing

// ENGINE 04 — CLAIM VERIFICATION

Cross-reference the claim.

Claude extracts factual claims from the post text. Each claim is cross-referenced against Google Fact Check Tools (IFCN-certified fact-checkers) and ClaimBuster's claim-worthiness model. Verdicts cite the fact-checker by name.

Claude · Google Fact Check API · ClaimBuster

// ENGINE 05 — THREAT INTELLIGENCE

Fingerprint the campaign.

Narrative, network, temporal, and behavioural fingerprints get matched against known disinfo campaigns. Findings export as STIX 2.1 objects and serve via a TAXII 2 endpoint — drop them straight into your TIP or SIEM.

STIX 2.1 · TAXII 2 · Campaign Similarity Score · TLP

// COMPOSITE TRUST SCORE

One number. Sourced.

All five engines feed a composite 0–100 trust score with a categorical band — VERIFIED, CREDIBLE, CAUTION, SUSPICIOUS, or THREAT. Never binary. Always sourced. Fully explainable in the linked evidence chain.