Home Products Truth Shield
Product · Verify what’s true

Truth Shield

Credibility analysis on public posts with the reasoning shown — never a black-box score. The core is open source, free for the communities that need it most.

A real claim analyzed live on the free tool: a fabricated vaccine-microchip conspiracy scores 95/100 CRITICAL, with the specific signals and reasoning shown — not just a number.

The problem it solves

Disinformation moves faster than fact-checking, and opaque “trust scores” make things worse — a number nobody can interrogate is just another authority to distrust. Truth Shield takes the opposite bet: show the reasoning, every time.

How it works

Truth Shield analyzes public posts for credibility signals and produces an assessment with its reasoning displayed — what it weighed, what it found, why it concluded what it concluded. A reader can disagree with the analysis, because they can actually see it. That is the point.

Open source, on purpose

The core is open source and free — a deliberate decision that verification tooling for vulnerable communities should not sit behind a paywall. The commercial suite builds on it; the foundation belongs to everyone.

Verified behaviors

Every claim below maps to a dated behavioral proof against the live production service.

A genuinely verified post displays as verified in the feed, not silently mislabeled by a display mismatch.
method: a frontend/backend verdict-string mismatch was found and fixed directly on the live system, then confirmed against a real analysis call · observed: a live control test ("water boils at 100°C") returned verdict verified, confidence 99, and the corrected display logic now renders it correctly · VERIFIED July 28, 2026
Ordinary, low-stakes content is not penalized with an elevated risk score just for lacking a citation.
method: the analysis prompt was found to have no carve-out for mundane claims; corrected and redeployed, then re-tested against the same content before and after · observed: a mundane local news claim scored confidence 25/unverified/moderate before the fix and confidence 15/opinion/clean after — control cases (a verified scientific fact, an actual disinformation claim) held steady, confirming the fix targeted the right case without weakening real detection · VERIFIED July 28, 2026

Honest limitations

Download the white paper (PDF)

Evaluate it against your platform

Book a demo

Where this fits

Candor works at the intersection of gaming safety, age assurance, privacy, parental consent, cryptographic evidence, and auditable trust & safety decisions. Related: