Proof of personhood, also called proof of humanity, proof of human or human verification, establishes that a real human being, often a unique one, stands behind an account, credential or interaction. Proof of Actor is broader: it asks who or what is actually performing a specific interaction, whether a human is present and in control right now, whether AI is participating, and, if an AI agent is acting, who authorised it and within what limits. Proof of personhood answers "are you a real human?". Proof of Actor adds "are you here, are you in control, and who is accountable for this action?"
The two are complementary, not competing. This essay explains what personhood systems do well, where their question ends, and how the Actor question extends it.
What proof of personhood means
The term covers a family of approaches with a shared goal: to let a service establish that a participant corresponds to a real human, without necessarily learning who that human is.
Two related properties are often bundled together:
- Humanity: the participant is a person, not a script or an AI system.
- Uniqueness: each person holds only one credential or account in a given system. This is what makes personhood useful against Sybil attacks, where one party creates many identities to gain disproportionate influence, rewards or votes.
Approaches to proof of personhood vary widely, and each makes different trade-offs:
| Approach family | General idea | Typical trade-offs |
|---|---|---|
| Document-based credentials | Derive a personhood credential from verified identity documents | Strong assurance, but depends on document access and raises data-handling questions |
| Biometric uniqueness | Use a biometric to establish that one person enrols once | Strong uniqueness properties, with significant privacy and consent considerations |
| Social or web-of-trust vouching | Existing members vouch for new members | Decentralised, but can be slow to bootstrap and open to collusion |
| In-person or event-based verification | Attendance at a physical or synchronous event | Good uniqueness, limited scale and convenience |
| Interaction-based human verification | Challenges or signals that indicate a human is interacting | Low friction, but typically weaker on uniqueness and increasingly tested by capable AI |
Many systems combine these with privacy-preserving cryptography so that a service can learn "this is a unique verified human" without learning name, location or other identifying data. That is an important design direction, and one we share.
We think the people working on personhood are addressing a real and growing need. As AI-generated content and automated accounts become cheaper to produce, the ability to say "a real person is behind this" has clear value for communities, social networks, polls, reviews, promotions and any system that allocates something per person.
Where the personhood question ends
Proof of personhood is typically established at a point in time: at enrolment, at account creation, or when a credential is presented. It answers a question about the existence of a human behind an identifier.
That leaves several questions open, and they are increasingly the ones relying services care about.
1. Is the human present now?
A credential presented at the start of a session tells a service that a verified human was involved then. It does not by itself establish that the same human is present forty minutes later when a payment is approved or an assessment is submitted. Presence is a property of the interaction, not of the identity.
2. Is the human actually in control?
A verified person can be present at a device while something else operates the session: remote-access software, another person at the keyboard, or an AI agent driving the browser. Presence and control are different. A person sitting in front of a screen is not necessarily the one directing what happens on it.
3. Is AI participating?
A verified human can use an AI assistant to write every message, answer every question or generate every review. The account is human, the person is real, and the credential is valid. Whether that matters depends entirely on context. For a casual message it usually does not. For a skills assessment, a sworn statement or a "verified human review" badge, it often does.
4. If an agent is acting, who authorised it?
Personhood systems are about humans. But a growing share of legitimate activity will come from AI agents acting for people and organisations. A merchant receiving an order from a shopping agent does not need the agent to be human. It needs to know which principal stands behind the agent, what the agent was authorised to do and whether this purchase falls inside that authority. That is a question about delegation, not personhood.
5. Who is accountable?
Ultimately, a service wants to know who stands behind an action. For a direct human action, that is the person. For an agent action, it is the principal who delegated authority. Personhood can contribute evidence to the first case, but it is not designed to describe the second.
Proof of Human vs Proof of Actor: a comparison
| Question | Proof of personhood | Proof of Actor |
|---|---|---|
| Is there a real human behind this? | Yes, core purpose | Yes, via the Personhood layer |
| Is this a unique human? | Often a core purpose | Supported where the use case needs it |
| Who is the human? | Usually deliberately not disclosed | Optional; "verified human, identity undisclosed" is supported |
| Is the human present during this interaction? | Generally out of scope | Presence layer |
| Is the same Actor continuing throughout the session? | Generally out of scope | Actor continuity layer |
| Is the human controlling the interaction? | Generally out of scope | Human control layer |
| Is AI materially participating? | Generally out of scope | AI participation layer |
| If an agent acts, who authorised it and with what scope? | Out of scope | Delegation and scope of authority layers |
| Did the human approve this specific action? | Out of scope | Intent layer |
| Who is accountable? | Partially, for direct human actions | Accountability layer |
The table is not a scorecard. Personhood systems are designed to answer their question well, and the answer is useful. The point is that the Actor question has more layers, and different services need different subsets of them.
How the two fit together
Within Proof of Actor, the personhood question is not discarded. It becomes one primitive among three.
- Proof of Human asks: is a real human here? It covers personhood and presence. It could consume evidence from a range of sources, including privacy-preserving personhood credentials where a service or user already has them.
- Proof of Control asks: is the human actually controlling this interaction? It covers continuity, human control and AI participation across a session.
- Proof of Authority asks: if an AI agent is acting, who authorised it and what may it do? It covers delegation, scope, limits, expiry and revocation.
The output in each case is designed to be an Actor Attestation, a machine-readable record of what could be established, with a confidence level rather than an absolute verdict. The relying service then applies its own policy.
A useful mental model is the path we use internally:
Proof of Human
|
v
Proof of Control
|
v
Proof of Actor
|
v
Proof of Authority
Personhood establishes that a human exists. Control establishes that the human is directing the interaction. The Actor view brings in AI participation and agents. Authority extends trust to agents acting on someone's behalf.
Examples: when personhood is enough, and when it is not
A community poll. The main concern is one vote per person. A unique-human credential may be exactly what the platform needs. Current presence and control matter less, and the platform may prefer not to learn identity at all.
A product review platform. "A real human wrote this" is useful, but a platform may also want to know whether the review text was generated by AI, and whether the reviewer actually purchased the item. Personhood is one input; AI participation and other context matter too. See reviews.
A dating service. Members want evidence that the person behind a profile is real. Over time, they may also want evidence that the person they are talking with is the person who verified, and that the conversation is not being conducted by an automated system.
A remote hiring assessment. A verified human candidate is the starting point, not the finish. The employer wants evidence that the same candidate remained present and in control, and whether AI materially participated. See hiring.
A checkout receiving agent traffic. The buyer may not be a human at all, and that can be perfectly fine. What matters is the principal, the mandate and whether this action is inside it. Personhood alone has nothing to say here.
Illustrative trust receipts
To make the difference concrete, here are two illustrative Actor Trust Receipts. These are conceptual examples from our Developer Preview design work, not output from a live system.
Proposed format (illustrative)
ACTOR TRUST RECEIPT
Actor Human
Personhood Verified
Identity Undisclosed
Presence Continuous
Human control High confidence
AI involvement Declared assistance
Session continuity High confidence
ACTOR TRUST RECEIPT
Actor AI Agent
Agent identity Verified
Principal Verified individual
Authority Valid mandate
Requested action Inside delegated scope
Human approval Required, received
A pure personhood result would correspond to roughly one line of the first receipt. The rest is what the Actor question adds.
Privacy considerations
Both personhood systems and Proof of Actor face the same responsibility: proving what matters without revealing what does not. We think the right default is selective disclosure. A forum may need "unique real human". A merchant may need "agent authorised to spend up to $500". A hiring platform may need "the same verified candidate remained present". None of them necessarily needs the underlying personal data.
Proof of Actor is being designed around that principle, and around collecting the minimum information required for the claim a service actually asks for.
Frequently asked questions
What is proof of personhood?
Proof of personhood is a way of establishing that a participant in a digital system is a real, often unique, human being, usually without revealing who they are. It is commonly used to resist Sybil attacks, where one party creates many identities to gain unfair influence.
Is proof of humanity the same as proof of personhood?
The terms are generally used interchangeably. Both refer to establishing that a real human, and often a unique one, is behind an account or credential. "Proof of human" and "human verification" are also used for related ideas, sometimes with a focus on interaction-time checks rather than one-off enrolment.
What is the difference between human verification and identity verification?
Human verification establishes that a participant is a person. Identity verification establishes which person they are, typically by checking documents or official records. A service can need one without the other: many communities need to know members are human without knowing their legal identity.
Can a verified human still be using an AI agent?
Yes. Having verified that a human exists says nothing on its own about whether that human is present, in control of the current session, using AI assistance or has delegated the action to an agent. Those are the questions Proof of Actor is designed to address. See What Is Proof of Actor?.
Building on personhood
Proof of personhood is an important foundation for a more trustworthy internet. As humans and AI agents increasingly share the same services, the next layer will need to describe not only that a human exists, but who or what is acting in each interaction and who is accountable for it. We are building that layer and refining it with early partners. If your service needs more than a single human check, join the Developer Preview or become a design partner.
