Verdict Schema

In one sentence

The verdict your backend reads is { country, confidence, alarm } — three fields, nothing about how they were derived.

Success response

Returned by GET /v1/verdict/{sessionRef} once a verdict is ready:

{
  "country": "DE",
  "confidence": 0.91,
  "alarm": "none"
}

Field	Type	Range / values	Meaning
country	string	ISO 3166-1 alpha-2 (e.g. "DE", "US")	Estimated country of origin for the session. May be empty if there wasn't enough signal.
confidence	number	0–1	Confidence in country. Higher is more confident.
alarm	string	"none" | "low" | "medium" | "high"	Escalation indicator for the session. Higher warrants more caution.

Supported surface

country, confidence, and alarm are the stable, supported fields. Treat anything else that may appear in a response as internal and unsupported — it can change or be removed. Build your policy only on these three.

Pending response

If the verdict isn't ready (the browser is still collecting, or no signals arrived for that sessionRef):

{ "status": "pending", "ref": "sess_abc123" }

returned with HTTP 404. Use the waitMs long-poll, or retry shortly. See Fetch the Verdict.

How to interpret it

• confidence gauges how much to trust country; choose a threshold for your risk tolerance and treat low confidence as "not enough signal", not as a negative.
• alarm is a separate axis — use it to decide how cautious to be, independent of country.

There is, by design, no field telling you why. See Verdicts and Trust & Privacy.

Where to go next

• Fetch the Verdict. How to read it.
• Verdicts. The conceptual read.