Facial age estimation
Check whether someone is old enough — from a selfie, in under a second, without asking for an ID up front. Estimation clears the obvious cases fast and privately; only users near your threshold need to show a document.
How it works
Zanyara analyses a selfie and returns an estimated age range with a confidence signal. You set the threshold (18, 25, or your own) and a buffer around it. Users comfortably above the buffer pass instantly; users inside the buffer are stepped up to document and NFC verification with liveness to confirm date of birth. No identity document is collected for the estimation itself.
Why estimation-plus-step-up
- Fast and private — most users never show an ID; lower friction, less data collected.
- Honest about limits — estimation gives a range, not a birth date, so borderline users escalate.
- Configurable — thresholds, buffers and step-up rules per use case.
- Ofcom-aware — facial age estimation is among the methods Ofcom lists as capable of being highly effective age assurance.
Estimation vs verificationAge estimation predicts a range from a face; age verification confirms an exact date of birth from a document or authoritative source. Most flows combine them. Read the difference →
Frequently asked questions
- How accurate is facial age estimation?
- Facial age estimation predicts an age range from a selfie, not an exact date of birth. It’s highly effective for people clearly above or below a threshold, and less certain for those close to it. Zanyara handles that honestly: set a buffer around your threshold and step borderline users up to document verification.
- Is it privacy-friendly?
- Yes. Estimation works from a selfie alone — no identity document required for the age check itself — which is why Ofcom lists facial age estimation among methods that can be highly effective age assurance. You control retention of the image; only step-up cases collect a document.
- Does it count as 'highly effective age assurance' under the Online Safety Act?
- Ofcom names facial age estimation as one of several methods capable of being highly effective, judged on accuracy, robustness, reliability and fairness. Whether a specific deployment meets the bar depends on configuration and your risk context — we help you set thresholds and step-up rules, but you own the compliance decision.