BulletPROOF

How BulletPROOF Works

From headline to Proof Score in minutes. We extract claims, find corroboration, analyze tone, and publish transparent reports—so readers and newsrooms can verify before they amplify.

Visual Overview
Ingest Extract Corroborate BiasLens™ Proof Score
Article goes in → claims come out → evidence checked → tone analyzed → Proof Score + transparent report.

The Workflow

Five stages. Built for speed, explainability, and accuracy.

Step 1

Ingest & Normalize

Paste a link, upload a PDF, or drop raw text. We extract main content, remove boilerplate, and detect the language and structure.

  • HTML/AMP/PDF; basic OCR for images*
  • Canonicalization & de‑duplication
  • Metadata capture (authors, timestamps)
Step 2

Claim Extraction

NLP identifies verifiable statements, entities, numbers, and quotes, mapping each claim to its supporting sentences.

  • Claim salience & check‑worthiness
  • Entity linking & numeric grounding
  • Quote attribution & stance
Step 3

Corroboration

We retrieve diverse sources and run stance analysis to see if each claim is supported, contradicted, or uncertain.

  • Freshness‑weighted retrieval
  • Outlier detection & contradictions
  • Confidence intervals
Step 4

BiasLens™ & Rhetoric

We analyze tone, framing, hedging, and sentiment to highlight where language departs from evidence.

  • Subjectivity & valence overlays
  • Loaded language & weasel words
  • Sourcing balance indicators
Step 5

Proof Score & Reports

Per‑claim assessments roll up into an article‑level Proof Score with an explanation and export options.

  • Explainable scorecards with evidence links
  • Configurable weights (recency, redundancy, source quality)
  • Export via ProofStack™ (PDF, CSV, JSON)

*Advanced OCR and media forensics via SignalGuard™ in higher tiers.

Under the Hood

An explainable pipeline—modular, auditable, and fast.

MultiProof™

Run multiple models and data sources concurrently; compare stances; compute consensus‑weighted scores.

  • Cross‑model agreement
  • Override rules & audit trails
  • Custom ensembles in higher tiers

BetterAngle™

Suggests stronger or more current sources and reduces echo‑chamber effects.

  • Recency & redundancy scoring
  • Outlet diversity constraints
  • One‑click citation swap

BiasLens™

Inline overlays for tone, sentiment, and framing—without changing the underlying facts.

  • Section heatmaps
  • Draft vs. published comparison

Private Model Zones™

Isolated routing with organization‑owned keys, prompts, and policy guards.

  • Zero data retention
  • KMS‑managed secrets
  • Per‑org usage caps

Certify™

Attach signed verification manifests readers can independently validate.

  • Public badges & linkback
  • Regenerate on updates (TruthTrail™)

ProofStack™

Export full reports and evidence bundles for archives and audits.

  • PDF, CSV, JSON
  • Bulk exports & redaction

API Example

SDK: JS • Python • Go
// Submit an article for verification
POST /v1/verify
Content-Type: application/json

{
  "url": "https://news.example.com/article.html",
  "options": { "biasLens": true, "export": ["pdf","json"], "model": "balanced" }
}

// Retrieve a report
GET /v1/reports/{report_id}

Webhooks and signed manifests available. Region‑locked processing for enterprise.

Integrations & Workflow

Works where you do—browsers, CMS, and code.

Browser

Highlight claims on-page, preview Proof Scores, and export to ProofStack™.

CMS

Pre‑publish checks with status badges; WordPress/Ghost hooks; StoryGuard™ webhooks.

Data

ProofStack™ bundles (PDF/CSV/JSON) and signed manifests via Certify™.

Trust & Transparency

What Scores Mean

Scores summarize evidence as of retrieval time—never an appeal to authority. Every score links to evidentiary context.

  • Supported / Contradicted / Uncertain stances
  • Uncertainty bands with confidence levels
  • Rationale & per‑claim breakdowns

How We Handle Change

Facts evolve. StoryGuard™ watches for updates; TruthTrail™ tracks versions; Certify™ regenerates manifests on change.

Examples

Sample results (fictional data):

“Coffee cures insomnia”

Proof Score 34%

Claim contradicted by clinical summaries; headlines overstate evidence.

“EV sales drop 40% worldwide”

Proof Score 59%

Partial support in one region; global numbers inconsistent or out‑of‑date.

“New vaccine reduces risk by 85%”

Proof Score 84%

Strong alignment with peer‑reviewed summaries and official data.

FAQ

Does BulletPROOF replace human fact‑checkers?
No. It accelerates research and makes evidence transparent. Editors still decide what to publish.
What if there are no prior fact‑checks?
BulletPROOF corroborates claims using diverse sources and stance analysis—even when no prior checks exist.
Can I export the full report?
Yes—use ProofStack™ to export PDF/CSV/JSON and share manifests with Certify™.
What languages are supported?
English primary, with multilingual retrieval/stance available in Pro/Enterprise pilots.

Verify before you amplify.

Try BulletPROOF on your next article and export a fully transparent report.

Get Started