Specificity: what it measures and what it cannot prove.
Measure whether text contains named entities, numbers, source cues, concrete examples, specific verbs, temporal detail, and geographic specificity.
Plain-English method
Specificity is a deterministic semantic-concreteness signal. It rewards claims that include names, numbers, dates, locations, examples, and source cues, then penalizes generic generalizations and vague wording.
Mechanism and scoring
The score is a weighted feature pass over parsed sentences and spans. It creates a generic weakened version to show how much meaning collapses when specifics are removed. This is useful for routing and revision, not truth verification.
What this catches
- Named examples and entities
- Numbers and concrete details
- Source and quote cues
- Vague claims that should be revised
What this misses
- Whether a claim is true
- Whether a source supports the claim
- Whether a human or AI wrote the text
This is one signal in a layered stack.
Single-method detectors are too easy to overtrust. Specificity is useful when it changes routing: allow, revise, human_review, or reject. It should be layered with specificity, provenance, pattern pressure, Unicode sanitation, media provenance, and paid Deep Scan when the decision matters.
check_specificity is available through local MCP with no LLM cost, and through the remote MCP endpoint with free unauthenticated rate limits. See all detection methodologies and the dedicated methodology page for the deepest treatment of this signal.