Find incorrect, partial and unsupported descriptions before treating visibility as success.
An answer can mention a business and still misstate its offer, location, status or evidence. Verify Presence separates appearance from factual accuracy.
How the workflow works
The scope stays bounded from observation through remeasurement.
- Confirm public evidence before analysis.
- Classify statements as correct, partial, incorrect or unresolved.
- Record contradictions between sources.
- Route material corrections to an accountable reviewer.
What the customer receives
Every output remains attributable to its source and period.
- Accuracy register
- Contradiction log
- Source-level evidence links
- Exact drafted corrections
Human-review boundary
Configured tools can support collection, comparison and drafting; controlled launch may use disclosed human assistance. Material facts, external publication and insufficiently evidenced changes remain human decisions.
- Confirm evidence before it informs diagnosis.
- Approve an exact version, not a moving draft.
- Revoke access or sharing when the authority ends.
Measurement limit
The product does not prove every public statement is true; the authorised reviewer remains responsible for confirmation.
- Small prompt sets are directional.
- AI answers can vary between runs.
- No ranking, recommendation or revenue guarantee.
What to know before you begin
What is fixed for comparison?
The prompt wording, target scope, measurement date, provider/model context and classification method are recorded. Material scope changes create a new version.
How is routine work delivered?
Configured tools can support scheduled collection, evidence matching, drafting and comparable remeasurement within configured permissions. Controlled launch may use disclosed human assistance, and sensitive changes retain human approval.
Does a better result prove the intervention caused it?
No. A measured difference is recorded with its limitations and does not automatically establish causation.