MediaMeter is one quality layer with three capabilities. They share the same evidence store, the same audit trail, and the same rule: AI advises, humans decide.
Three capabilities, one record
Assess
Seven dimensions, weighted by content type, with evidence attached to every score.
Read the assessment spec
Fact-check
Claim extraction, evidence retrieval, and IFCN-aligned verdicts with a preserved chain.
Read the fact-check spec
Detect
Text, image, audio, and video signals combined with C2PA and SynthID provenance checks.
Read the detection spec
What makes it different
- Evidence first. A score without a citation is a guess. Every dimension links to the passage, source, or signal that produced it.
- Human review where it matters. FALSE, MISLEADING, and AI_GENERATED verdicts require a named reviewer before they are published.
- Built to be audited. Model versions, prompts, and reviewer decisions are recorded so an assessment can be reconstructed months later.
- No auto-publishing. MediaMeter never suppresses, ranks, or publishes content. It produces findings for people to act on.
Where it fits
Newsrooms use assessment to make quality measurable across desks. Fact-checking teams use the evidence chain to shorten verification without losing rigour. Broadcast and social teams use detection before material goes to air or feed. Integrations bring the same checks into WordPress, Ghost, Contentful, Arc XP, Slack, and the browser.
The capabilities described here are a design specification. Nothing on this site represents a shipped integration, a certification, or a paying customer.