RESEARCH AND METHODOLOGY
Why the check works this way
The product uses short-lived challenge-response recording to make a prepared clip harder to reuse. It combines automated signals with direct human review because neither layer is reliable enough on its own.
Seven seconds is a product format, not a truth test
Research shows that people form impressions quickly and that short video excerpts can contain useful social information. It does not show that seven seconds proves identity, honesty, or safety. Seven seconds was chosen to keep the request practical while leaving time for a fresh action and spoken phrase.
Fresh actions reduce simple replay
NIST describes printed photos, displayed images, and replayed videos as presentation attacks. A prompt issued shortly before recording makes a prepared response less useful, especially when it combines motion, speech, face framing, and a hand action.
Automated checks are deliberately advisory
Camera models can miss genuine blinks, gestures, accents, or low-light recordings. They can also be manipulated. The service therefore labels signals as strong, partial, or inconclusive and releases the complete submitted attempt history for the requester to inspect.
Capture details are context, not identity
IP address, timezone, browser headers, screen size, and camera settings may help the requester notice inconsistencies. They do not prove physical location or identity and can be changed by proxies, VPNs, privacy tools, virtual devices, or a modified client.
Why the useful intervention point is before trust becomes expensive
The US Federal Trade Commission reported $3.5 billion in consumer losses to imposter scams during 2025, with nearly one in three fraud reports involving an imposter. It also reported $2.1 billion in losses from scams that started on social media during 2025; nearly 60% of reported romance-scam losses began on a social platform.
These figures do not show that a Human Check prevents scams. They support the product decision to create a deliberate pause, use a second channel, and add fresh evidence before money, documents, travel, or account access becomes difficult to recover.
Known limits
A browser-only check is not deepfake-proof and cannot detect coercion, guarantee that an account belongs to the person in the video, or establish that the person is safe. The result should support a decision, not replace other verification or safety precautions.