Age estimation

Predicting someone's rough age from a signal like a selfie, voice sample, or account activity, most often a selfie, without collecting an ID.

For example, a person takes a quick selfie, and the system returns an estimated age or confirms they are comfortably above the required age, with no document involved.

Model-based inference of an age range from a biometric or behavioral signal, most commonly facial imagery, evaluated against a challenge-age threshold. A challenge age is a cutoff set above the legal age, so anyone estimated close to the line gets a second check.

A typical facial age estimation flow works like this:

1. The person takes a selfie, usually with a liveness check to confirm a real person is present.
2. A model analyzes the image and outputs an estimated age or age range.
3. The service compares the estimate with its challenge age.
4. People estimated safely above the line pass. People near or below it are sent to another method, such as an ID check.

The signal can come from different places:

• Facial imagery is the most common source.
• Voice samples can also support an estimate.
• Behavioral signals, such as account activity, can add context.

Age estimation sits next to terms that are easy to mix up:

• Age verification establishes age from a verified document or record, while estimation predicts it.
• Face match compares one face with another to confirm they are the same person, while age estimation analyzes a face only for age.

For more, see Age Assurance Explained: Verification, Estimation, Segmentation, and Gating