TrustMark 5D

Five dimensions. One number you can take apart.

TrustMark™ never collapses its reasoning. Each dimension is scored on its own, weighted into the total, and stays inspectable long after the decision is made.

See it on a live claim →How a score is built

Sample assessmentIllustrative

85TrustMark scoreDefensible

D1Truth Alignment91D2Source Integrity84D3Author Credibility78D4Rhetorical Neutrality88D5Visual Authenticity72

The five dimensions

Each one scored alone, before anything is combined.

Weighted composition100%

D1 · 35%D2 · 25%D3 · 15%D4 · 15%D5 · 10%

The weights ship with the score — never buried in the model.

D1w 0.35

Truth Alignment

Claim verification grounded in web search and fact-checks — extracts atomic claims and scores them against the evidence base.

Atomic claim extractionFact-check signalsContradiction detection

35% of the composite

D2w 0.25

Source Integrity

Per-source credibility shown alongside every verdict — domain reputation, transparency, and historical accuracy by tier.

Domain tiering T1–T4Transparency signalsHistorical accuracy

25% of the composite

D3w 0.15

Author Credibility

Weighs the person behind the content — historical accuracy, domain expertise, and consistency over time.

Domain expertiseTrack recordConsistency over time

15% of the composite

D4w 0.15

Rhetorical Neutrality

Detects emotional intensity, urgency triggers, and persuasion tactics. Misinformation is often persuasive, not false.

Emotional intensityUrgency triggersPersuasion tactics

15% of the composite

D5w 0.10

Visual Authenticity

Verifies images and video via reverse-image matching, manipulation probability, and metadata consistency.

Reverse-image matchingManipulation probabilityMetadata consistency

10% of the composite

Reading the number.

A score is a recommendation about how much weight to put on something — not a verdict on whether it is true.

70–100Defensible

Cite it, act on it, and be able to show the reasoning if you are asked to.

40–69Qualified

Usable with the caveat attached — the dimension dragging it down is named for you.

0–39Contested

Do not rely on it alone. The proof chain shows exactly which evidence failed.

The proof chain

Every step is recorded, not inferred.

The same input always produces the same explainable result. Re-run a claim next month and you can see exactly what changed, and why.

See it on a live claim →

Proof chain · one claima3f72c9b1e4d…

✓Claim extraction1 atomic claim · LLM

✓Web search grounding12 results · 3 Tier 1 sources

✓Fact-check · LLM signalsVerified · confidence 0.94

✓D1 scoring · Truth Alignment91 · weight 0.35 · PSD §1.5

✓TrustMark persistedCRV updated · audit trail written

Early access

ScoRISK is in private preview.

Two required fields to get in the queue — the rest is optional and helps us tailor your access.

No spam, ever — we review every requestAll five TrustMark dimensions, liveEarly access shapes what we build next

Flexible credit-based pricing — details at early access.