Proof of personhood is a way for a website, app or network to know that an account belongs to a real, unique human being, and that the same person isn't behind many accounts. Done well, it does this without the site learning who the person is.
It answers two questions that the internet was never built to answer:
- Is there a real person here, or a bot, a script or an AI agent?
- Is it a different person from the other accounts, or one person with many?
The second question is the hard one. Plenty of tools can say "probably not a bot". Very few can say "one person, one account".
Why it matters now
Creating an account costs nothing, and AI has made fake accounts cheaper and more convincing than ever. One person can run hundreds of accounts that write like humans, pass CAPTCHAs and look real. That breaks anything that assumes one account means one person:
- Online communities: sock puppets, brigading, and banned users coming straight back with a new account.
- Free tiers and trials: the same person signing up again and again.
- Sign-up bonuses, referrals and giveaways: one person farming many rewards.
- Polls, votes and reviews: ballot stuffing and fake reviews.
- Marketplaces and dating: scammers who return under a new name after every ban.
In security this is called a Sybil attack: one actor pretending to be many. Proof of personhood is the defense against it.
A short history of the idea
- 2017: the term. Researchers at EPFL in Switzerland proposed "proof-of-personhood" in a paper on fairer cryptocurrencies. Their method was the pseudonym party: people gather in person at the same time, each receives one anonymous token, and nobody can be in two places at once. The goal was one person, one vote, with anonymity intact.
- Late 2010s to early 2020s: crypto. Airdrops, on-chain voting and universal basic income experiments suffered badly from fake accounts, so projects tried social vouching, video verification parties and biometric hardware.
- 2023: the biometrics debate. Ethereum co-founder Vitalik Buterin wrote that proof of personhood is valuable, that a world without it would likely drift toward centralized identity, and that no single approach is ideal. He set out privacy, accessibility, centralization and security as the risks to weigh.
- 2024: mainstream. A paper by 32 researchers from OpenAI, Microsoft, Harvard, MIT, Oxford and other institutions proposed personhood credentials: privacy-preserving credentials that let people show they're human online without revealing who they are, issued by governments or other trusted institutions, and not necessarily using biometrics. The problem had moved from crypto to the whole internet.
See personhood credentials for more on that proposal.
How it works
Most proof-of-personhood systems follow the same four steps, whatever they use underneath:
- Enroll. The person proves they're real once: with a government ID, a biometric, in-person attendance, or vouching from others.
- Check uniqueness. The system makes sure this person hasn't enrolled before, by comparing something that's one per person: a document, a biometric, a physical presence.
- Issue something reusable. The person gets a credential, account or key they can use again without repeating the check.
- Prove it to sites, privately. Each site learns "this is a unique, verified person" and gets an identifier for that person. In privacy-preserving designs, that identifier is different on every site, so sites can't link people.
The differences between systems come down to what they use in steps 1 and 2, and how much each party learns in step 4.
The main approaches
There's no perfect method. Each trades off how strongly it proves uniqueness, how private it is, how many people can use it, and what it costs.
| Approach | Proves a unique person? | Privacy | Who can use it |
|---|---|---|---|
| CAPTCHAs | No: only "probably not a simple bot" | Good | Almost everyone |
| Email or phone verification | Weakly: numbers and addresses are cheap to get | Fair | Almost everyone |
| Payment card checks | Weakly: cards can be shared or bought | Poor: the site learns payment details | People with cards |
| Device attestation | No: proves a real device, not a unique person | Good | Recent devices |
| Social vouching (webs of trust) | Partly: depends on the honesty of the network | Varies | People in the network |
| In-person events (pseudonym parties) | Yes, for those who attend | Good | People who can attend |
| Iris or other biometrics | Yes | Biometric data is collected, even if processed carefully | People who can reach the scanning hardware |
| Government ID checks (traditional KYC) | Yes | Poor: the site usually receives and stores the ID | People with a government ID |
| Government digital wallets and mobile licences | Not on their own: they share attributes, not uniqueness | Good | Growing, but limited by country and state |
| Privacy-preserving ID-based checks | Yes | Good: sites get only a yes and a per-site ID | People with a government ID |
Two comparisons come up most often:
- Proof of personhood vs KYC. KYC proves who someone is and hands that identity to the business. Proof of personhood proves that someone is a unique person. See proof of personhood vs KYC.
- Biometric vs document-based. Biometrics are one per person everywhere but raise privacy and access concerns; documents are familiar and widely held but vary by country. See biometric proof of personhood.
What good proof of personhood looks like
A system worth trusting should:
- Prove uniqueness, not just "human". One person should get one account per site.
- Share nothing else. The site needs a yes and a stable ID for that person, not their name, birth date or document.
- Stay unlinkable across sites. The same person should look like a different, unrelated ID on every site, so sites can't combine what they know.
- Be usable by ordinary people, with things they already have, without special hardware or a crypto wallet.
- Handle real life: lost access, renewed documents, mistakes, and people who need to leave. Recovery and deletion matter as much as enrollment.
- Be honest about who could link people. In some designs nobody can; in others, a trusted operator holds keys that could. Either can be reasonable, but users and sites should know which.
How Human Pass does it
Human Pass is a sign-in that proves a person is a real, unique human, and nothing else.
- Verify once. A person checks a government ID with an identity-verification provider: a photo of the document, plus a liveness and face-match check.
- Keep only fingerprints. Human Pass keeps one-way fingerprints of the document, never the document itself, the name or the birth date. If the same document is used again, it leads back to the same Human Pass, so one person can't hold two.
- Sign in anywhere. Sites add a "Continue with Human Pass" button (standard OpenID Connect). Each site gets a yes, "verified, unique person", and an ID for that person that means nothing anywhere else.
The site gets the benefit of an ID check, with nothing to store and nothing to leak. See how it works for sites and what people can expect.
Its limits, said plainly: it needs a government ID, which not everyone has. Someone holding qualifying IDs from two countries could in principle create two Human Passes; a check on name and birth date flags these for review. And Human Pass is a trusted middle party that could, with its keys, link a person's IDs across sites: our security overview explains how keys, data and access are locked down.
Common questions
Is proof of personhood the same as identity verification?
No. Identity verification proves who someone is, and usually hands that identity to the site. Proof of personhood proves that someone is a unique real person. A privacy-preserving system can use an ID check behind the scenes and still tell the site nothing about who the person is.
Is proof of personhood the same as proof of humanity?
They're used almost interchangeably. Both mean showing that an account belongs to a real, unique human. "Proof of Humanity" is also the name of one specific crypto project.
Does proof of personhood stop bots completely?
It stops one person or bot operator from creating many verified accounts, which is what makes bot farms work. A real person can still run a script from their own single account, but they only get one.
Does proof of personhood require biometrics?
No. Biometrics are one approach. Others use government documents, in-person events or social vouching. Many ID checks include a face match to confirm the document belongs to the person holding it, but the biometric data doesn't have to be kept.
Is proof of personhood only for crypto?
No. The term became popular in crypto, where airdrops and on-chain voting suffer badly from fake accounts, but the problem is everywhere: communities, marketplaces, free tiers, reviews and polls.
Can a site using proof of personhood track me?
In a privacy-preserving design, no: each site gets a different identifier for you and learns nothing else, so sites can't combine what they know. Check what a given system shares before you use it.
Further reading
- Borge, Kokoris-Kogias, Jovanovic, Gasser, Gailly and Ford, Proof-of-Personhood: Redemocratizing Permissionless Cryptocurrencies, IEEE European Symposium on Security and Privacy Workshops, 2017.
- Adler et al., Personhood credentials: Artificial intelligence and the value of privacy-preserving tools to distinguish who is real online, 2024.
- Proof of personhood on Wikipedia.