Does the asset match its own metadata?
Coherence is the public name for Context Consistency: whether what is visible in the work matches what its metadata claims about where, when, who, and what it shows.
What it reads
Coherence cross-checks the visual content against the file's own metadata claims. It is the only signal that compares two sources against each other rather than reading one in isolation.
Whether identifiable people, objects, brands, or entities visible in the work match what is named in its metadata.
Whether the environment, architecture, or geographic indicators visible in the work match the stated location.
Whether observable signals, lighting, season, technology, fashion, match the stated date.
Whether the production characteristics of the work match the stated creator, agency, or source.
Built-in guardrails
Checks content against metadata
A caption naming one city means nothing here until the visible content confirms it. Coherence is built specifically to catch the gap between claim and content.
Evaluates internal consistency
The check is internal: whether the work's own evidence agrees with itself.
Separate from metadata completeness
Custody asks whether metadata is present and complete. Coherence asks whether it is telling the truth. A file can pass one and fail the other.
Where it shows up
Legal and compliance
A Coherence contradiction is exactly the kind of finding an audit asks for and a Slack thread cannot produce.
See Custody
How Coherence and Custody split "is the metadata there" from "is the metadata true."
See the full formula
How Coherence combines with the other four signals into the single IQ Score.
See whether your metadata is telling the truth
We cross-check subject, location, timing, and creator against what's on screen.
Get early access