Skip to content
ScoRISKTrust IntelligenceRequest early accessRequest early access
← Back to blog
21 Jul 2026·2 min read·Product

Inside TrustMark™ 5D: How We Score Content Credibility

TrustMark doesn't give you a thumbs up or down — it shows every dimension, every source, and every reasoning step behind the score.

SScoRISKTrust Intelligence
TrustMark 5D scoring engine — dimension breakdown, evidence trail, and full audit chain behind a credibility verdict.

Most "fact-check" tools give you a thumbs up or down. TrustMark gives you a reason.

TrustMark™ is our credibility scoring engine. Instead of one opaque number, it breaks content down into independent dimensions, verifies claims against real evidence, and shows its work — every score traces back to the sources behind it.

A TrustMark claim credibility report: a score of 89, High Credibility, with confidence and evidence strength broken out.

Top features today

  • Truth Alignment scoring — every claim is checked against independent evidence, not just the subject's own website. A claim "verified" only by its manufacturer or author doesn't count as verified here.
  • Source Integrity scoring — we track the reputation of the domains behind your evidence: transparency, historical accuracy, and correction behavior, built from public reputation data (Wikipedia's reliable-sources list, academic research, and more) instead of guesswork.

The Truth Alignment and Source Integrity dimensions, each scored, weighted, and broken down into the checks and sources behind the number.

  • Full explainability chain — raw signals, sub-scores, and a plain-language reasoning narrative for every result. Nothing is a black box.
  • Audit trail — every score is versioned, timestamped, and reproducible. Re-run it later, get the same number, or see exactly what changed and why.

Raw evidence JSON behind a score — claim hash, source-integrity breakdown, and the full evidence trail, exported for reproducibility.

  • On-demand reverification — claims and sources can be re-checked as new evidence emerges, and the score updates accordingly.
  • Built-in skepticism — a claim doesn't flip to "verified" on a single ambiguous signal; the engine requires confirming evidence before it moves the needle, avoiding false swings.

Raw tool output logs from the verification pipeline — search, fact-check, fetch, sticky-replay, and LLM steps, each independently inspectable.

TrustMark isn't trying to tell you what to believe. It's trying to show you why — so you can check its work as easily as it checked the source's.

ProductRequest early access

Read next

All posts →