Glossary
The terms used in these docs, limited to what AdPriva runs today.
Traffic proofs
AdPriva Tag
A headless JavaScript tag that a publisher loads from https://cdn.adpriva.com/tag.js (older embeds: https://cdn.adpriva-ads.com/tag.js). It renders nothing and sets no cookies. On each page load it sends one view event with aggregated behavioural signals. It can also witness consent on forms the publisher marks. See AdPriva Tag.
Proof (Proof of Engagement)
The sealed record of one traffic event. The ingest server checks the event’s nonce, signs the event with an HMAC and stores it. The signature is the proof’s public id. See Proofs of Engagement.
View proof
The proof the tag sends for a page load (type: "view"), about one second after the page loads.
Nonce
A one-time random value inside every event. The server refuses an event whose nonce it has already seen, so a captured event cannot be replayed.
Validity score
A number from 0 to 100 given to each view proof. Higher means more likely to be a person.
Verdict
The label that goes with the validity score: human, uncertain or abusive_bot. The public verdict comes from the L0 (user agent) and L1 (IP class) checks only. The scoring service’s private L2 and L3 result can also be authorized_agent. See Proofs of Engagement.
Authorized agent
An automated client, such as a search crawler or an AI assistant, that proves who it is with a valid signed request (Web Bot Auth, RFC 9421). It gets its own verdict instead of being counted as a bot. Only the scoring service’s L3 check gives this verdict, and that result is private today, so it never appears as a public verdict.
Shadow mode
How AdPriva runs today. Every proof is scored, and its verdict is stored and published, but no proof is blocked or dropped because of its verdict. The scoring service’s behaviour (L2) and agent-identity (L3) checks also run in shadow mode: their result is kept privately and does not change the public verdict.
Anchoring and verification
Batch
A group of proofs anchored together. The anchor worker builds a new batch every 5 minutes from the proofs that are not yet in one.
Merkle root
One hash that commits to every leaf in a batch, so adding, removing or swapping a proof changes the root. For a view proof the leaf is the proof’s internal number, so the root fixes which proofs are in the batch, not their score, verdict or other fields. In a consent batch the leaves are the certificate signatures, which are computed over each certificate’s content, so the root also fixes that sealed content.
Merkle inclusion receipt
The leaf, sibling path and path indices for one proof. With it, anyone can recompute the batch’s Merkle root and compare it with the anchored value.
Zero-knowledge proof (ZKP)
A retired step. For view batches anchored before 2026-09-28, the anchor worker generated a Groth16 proof that a leaf belongs to the batch’s Merkle root, and zkVerify verified it before the root was anchored. The step was retired on 2026-10-16: new batches have no zero-knowledge proof and are anchored directly. See Legacy: zero-knowledge step.
Anchor
A transaction on the Horizen chain that calls AdPriva’s Anchor.sol contract and emits Anchored(merkleRoot). The contract stores nothing else, so the event is the public record.
Explorer
The public, read-only lookup at explorer.adpriva.com. You can search by proof id, batch handle or anchor transaction hash. It shows no publisher, site or visitor data. See Explorer.
Console
The signed-in dashboard at console.adpriva.com where an organization sees its own proofs, verdicts and traffic reports.
Portable Verification Certificate (PVC)
A signed JSON file that you download from the console’s Explorer page for a date range. It lists your organization’s scored proofs in that range by proof id, type and verdict, and carries a SHA-256 digest and an HMAC over its content. When all the proofs fall in one batch, it also names the batch’s Merkle root and anchor transaction.
Consent evidence
Consent certificate
Sealed evidence that a form marked data-adpriva-consent was submitted with a given consent wording and checkbox state. Anyone can look it up by its signature. See Consent Certificates — Quickstart.
Consent Evidence certificate (cert_v: 1)
The newer certificate for forms that declare a registered Legal Bundle (data-adpriva-bundle). It is signed with Ed25519 and anchored on Horizen, and it can be checked offline without contacting AdPriva. See Verifying a Consent Evidence Certificate.
Record-only certificate (cert_v: 2)
A Consent Evidence certificate for a consent form submitted without consent through the site’s banner, on a form the publisher opted in with data-adpriva-record-without-consent. It holds the same record as cert_v: 1 but no interaction signals, so its interaction assessment is NOT_AVAILABLE. See record-only evidence.
Legal Bundle
The registered, unchangeable set that a Consent Evidence certificate binds to, including the exact consent text. See Consent Evidence — Integration.
Subject reference (subject_ref)
The pseudonymous key a consent certificate is filed under. When the form supplies the publisher’s own user id, the tag sends a site-scoped SHA-256 hash of it, never the id itself. Publishers use it to answer access requests. See Data-Processor Posture.