The bot problem is now an infrastructure crisis. AI has made it trivially cheap to manufacture identity at scale: synthetic accounts, device farms, and coordinated fraud rings that look, behaviorally, almost indistinguishable from real users. Digital platforms can no longer assume that a verified account equals a real human.
Proof of humanity—the ability to cryptographically or probabilistically attest that an account is controlled by a unique, living person—has become one of the most consequential problems in fraud prevention.
Most proposed solutions reach for biometrics: scan an iris, match a face, issue a credential. It works, but it comes with real costs. High friction. Hardware dependency. A single point-in-time snapshot that creates a credential that ages the moment it's captured and can be easily handed over to another person or even a bot.
One basic example of this vulnerability is the driver re-verification issue in ride-hailing and food delivery apps: when prompted to take a selfie, inauthentic drivers return to the original driver's home to scan their faces, then return to work on the original driver's behalf.
We think there's a better path. And it starts with something we've been building for years: technology that is already embedded in over one billion devices worldwide.
|
1B+
|
18M
|
|
1 in 17M
|
1 in 1M
|
Incognia is already 55x larger in reach than the leading biometric alternative and 10x to 100x more accurate, with no hardware rollout and no user enrollment required.
Real humans have homes. They have routines. They commute, sleep, and run errands on weekends. Over weeks and months, these behaviors create a physical signature that is statistically unique and nearly impossible to fake at scale.
A fraud farm running 10,000 synthetic accounts cannot generate 10,000 distinct, internally consistent location histories going back months. A bot doesn't have a morning commute. An emulator doesn't leave the house.
Incognia's core technology has always been built around this insight. We analyze behavioral location signals to establish device identity and detect fraud. What we're now packaging is a direct output of that same signal stack: a Physical Presence Score,a continuous, privacy-preserving attestation that a device is held by a real, unique human moving through the real world.
A real person's movement through the world (home location, daily routine, commute, weekend patterns) creates a behavioral signature that is both unique and persistent. We measure this continuously, not as a point-in-time snapshot. The result is a False Acceptance Rate of 1 in 17 million, which is 10x to 100x better than iris recognition benchmarks measured by NIST: uniqueness at scale, without biometrics.
Our location graph clusters devices by physical co-presence. When 50 accounts share the same home location, that isn't 50 humans. It's a device farm. This signal alone eliminates entire categories of coordinated fraud.
GPS spoofing detection, emulator detection, and hardware attestation together answer a fundamental question: is this a real device, held by a real person, moving through the real world? Bots fail this test.
Unlike a biometric scan, our signal is continuous. A person's humanity score is maintained and refreshed, not issued once and forgotten. This means the attestation stays current, and anomalies surface immediately when behavior deviates from the established pattern.
| 1B+
|
1 in 17M
|
Continuous
|
The most prominent alternative in this space is World ID from Worldcoin, a system that uses a physical orb device to scan users' irises and issue a cryptographic proof of unique humanity. It addresses a real problem. But the tradeoffs are significant.
| Dimension | Incognia | World ID |
|---|---|---|
| User experience | Frictionless and continuous. No user action required | High friction. Requires in-person visit to an orb location |
| Scale and reach | Embedded in 1B+ devices globally | 18M users, limited to orb deployment locations |
| Accuracy | 1 in 17M false acceptance rate, 10x–100x better than iris (NIST) | Iris recognition benchmark: ~1 in 1M |
| Verification method | Behavioral location signals, device integrity, uniqueness | Iris biometric scan via proprietary hardware |
| Attestation type | Continuously refreshed score | Point-in-time credential |
| Hardware dependency | None | Requires physical orb device |
Proof of humanity shouldn't depend on whether there's an orb in your city.
The deeper issue with point-in-time biometric attestation is that it tells you who someone was at the moment they scanned, not who controls the account today. Credentials can be sold, transferred, or compromised. A continuous behavioral signal can't be handed off. It follows the person, not the token.
The Physical Presence Score is useful anywhere a platform needs to distinguish real users from synthetic ones and do it at scale, without adding friction. That includes fraud prevention for financial services, bot detection for consumer platforms, and increasingly, trust infrastructure for AI agent ecosystems where the question of human-in-the-loop accountability is becoming critical.
We've spent years building the underlying signal stack. The Humanity Score is how we're making it directly addressable as a product output.
Interested in integrating the Physical Presence Score into your trust stack? We're working with early partners now.
Get in touch with our team to learn more about the API and integration options.