HOW IT WORKS

A check that proves a real human — without learning who they are.

The fastest way to understand it is to take the check yourself.

THE PIPELINE

A check that runs where the respondent already is — their browser.

Step 01

Opens where they already are

The check loads in the respondent’s browser when they reach your survey. No app to install, no redirect maze, no friction.

Step 02

~5 seconds, on the device

Passive multi-signal analysis runs locally — face geometry (optional), micro-movement, device, and session. The camera feed never leaves the device.

Step 03

A signed verdict, pre-survey

You receive a signed result — real human · unique · not a bot — before they answer a single question. Fraud is stopped at the door, not cleaned up later.

Your device · in the browser

Camera on~5 seconds

Face geometry + behavioranalyzed locally

The image and video never leave this box.

non-reversible vector

Signed verdictreal · unique · not a bot

WHY IT TAKES A FEW SECONDS

Those seconds are the privacy guarantee.

Most fraud tools feel instant because they ship your respondent's data to a server — IP address, a device fingerprint, a selfie, sometimes a government ID. VerifyHuman does the opposite.

The analysis runs on the respondent's device. Those few seconds are real biometric and behavioral processing happening locally. What leaves the device is a 128-dimensional geometric signature that cannot be reversed into an image of your face — never a photo, and never your name or identity.

No image ever leaves the device
The camera feed is processed in-browser and discarded. Nothing is uploaded or stored.
Identity never learned
No name, email, or document, and we never use your IP as an identity. Only a non-reversible geometric signature.
Privacy-protective by design
No images or video are ever stored, we never learn the respondent’s identity, and we process only the minimal de-identified signals needed to detect fraud.

WHERE YOUR FACE GOES

VerifyHuman · on-device

Camera in your browser

Geometry computed locally

Only a vector → verdict

Your face never leaves your device. No image is uploaded or stored.

Server-side selfie · typical

Selfie captured

Full image uploaded

Stored & matched on a server

Your photo is sent to a server, where it is matched and may be retained.

TWO MODES

Biometric when you can use a camera. Behavioral when you can't.

Same SDK, same API. The camera is the single strongest fraud signal there is — reach for it whenever consent allows; fall back to behavioral-only when it doesn't.

Strongest signal

Biometric

The face is the one thing a fraudster can't swap. They can spin up a fresh device, a clean IP, and human-like behavior — and still fail here, because they can't become a different, live person on demand. Face geometry proves a real, present human and that they haven't already taken your study — without ever learning who they are.

· Liveness — a present person, not a photo, replay, or deepfake
· Uniqueness — catches the same human twice, even on a new device or IP
· Demographic check — flags a claimed age/gender that doesn't match (beta)

Non-biometric

Behavioral, device, and session signals only — no camera. For sensitive populations, low-friction flows, or contexts where camera consent isn't on the table. Still stops bots, click farms, and AI agents — biometric just adds the human-uniqueness layer on top when you can use it.

· Behavioral micro-signals + device + session
· AI-agent / LLM bot resistance
· Zero camera, zero biometrics

No trade-off

Strong checks, soft footprint.

Worried a camera costs you completes? It doesn't have to. VerifyHuman is passive-first — the check mostly just watches for a few seconds, no puzzles or hoops, and asks for more only when a respondent looks ambiguous. You can even run it camera-optional, so most respondents are cleared by behavioral signals alone and the camera steps in only for the cases worth a closer look. And nothing ever hard-stops a real person — if a check can't finish, it fails open and flags the response for review. You keep your response rate and your data quality.

DEMOGRAPHIC CHECK · BETA

Catch demographic fraud — on-device, in the same five seconds.

In biometric mode, VerifyHuman estimates an age range and gender from facial geometry, in the browser, so you can compare it against what the respondent claimed. It catches the 45-year-old who says he's a 24-year-old woman to qualify for a quota. Like everything else: derived from geometry, computed on-device, never stored, never an image.

Beta — estimates, not identity. A signal to compare, not a verdict.

Convinced it works? See what it costs.

Pay per verification, rate drops with volume — and an honest look at how we compare.

Pricing & comparison Sign up free