> ## Documentation Index
> Fetch the complete documentation index at: https://agenticintentprotocol.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Access & Citation Events

> Publisher → platform / operator telemetry for post-retrieval accountability (no ads, no pricing, no user identity).

## Purpose (normative)

Access and Citation Events provide **post-retrieval accountability** under the Agentic Intent Protocol.

They allow publishers to:

* record when content was accessed,
* verify how it was used,
* audit citations shown to users.

They allow platforms/operators to:

* prove compliant usage,
* reconcile settlement (outside this spec),
* support disputes.

Events are **append-only facts**, not signals. No ads. No pricing. No user identity leakage.

***

## Event types

v1.0 defines **two canonical event types**:

1. **`access`**
2. **`citation`**

Both share a common envelope.

***

## Canonical schema location

```
https://schemas.agenticintentprotocol.com/aip/1.0/event.schema.json
```

***

## Schema summary

| Field         | Type               | Required                        | Description                                                    |
| ------------- | ------------------ | ------------------------------- | -------------------------------------------------------------- |
| `aip_version` | string             | Yes                             | Must be `"1.0"`.                                               |
| `event_id`    | string             | Yes                             | Unique identifier for this event.                              |
| `event_type`  | string             | Yes                             | `access` or `citation`.                                        |
| `timestamp`   | string (date-time) | Yes                             | ISO 8601 when the event occurred.                              |
| `request_id`  | string             | Yes                             | `request_id` from the associated RetrieveRequest.              |
| `publisher`   | object             | Yes                             | `id`, `domain` (hostname).                                     |
| `platform`    | object             | Yes                             | `id` (platform that initiated the request).                    |
| `access`      | object             | When `event_type` is `access`   | `chunks_returned`, `token_count`, `retrieval_mode` (`chunks`). |
| `citation`    | object             | When `event_type` is `citation` | `source_url`, `chunk_ids`, `display_surface`.                  |
| `extensions`  | object             | No                              | Namespaced extensions.                                         |

***

## Semantics

### `access` event

Emitted **when content is successfully returned** in a `RetrieveResponse`.

It answers:

* was access granted?
* how much content was returned?
* under what retrieval mode?

It does **not** answer:

* how the content was ranked
* whether it was monetized
* whether the user clicked anything

### `citation` event

Emitted **when the publisher's content is surfaced to the user** as a citation.

It answers:

* which URL was cited
* which chunks were referenced
* where it was displayed

It does **not** imply endorsement, licensing, or commercial use.

***

## Required emission rules

* Publishers **MUST** emit an `access` event for every successful retrieval.
* Platforms **SHOULD** emit `citation` events when citations are shown.
* If `endpoints.event.required = true` in `aip.json`, platforms **MUST** emit citation events.

***

## Example: Access event

```json theme={null}
{
  "aip_version": "1.0",
  "event_id": "evt_access_001",
  "event_type": "access",
  "timestamp": "2025-11-14T18:22:01Z",
  "request_id": "req_92fA1",
  "publisher": {
    "id": "forbes",
    "domain": "forbes.com"
  },
  "platform": {
    "id": "openai_chat"
  },
  "access": {
    "chunks_returned": 1,
    "token_count": 18,
    "retrieval_mode": "chunks"
  }
}
```

***

## Example: Citation event

```json theme={null}
{
  "aip_version": "1.0",
  "event_id": "evt_citation_002",
  "event_type": "citation",
  "timestamp": "2025-11-14T18:22:05Z",
  "request_id": "req_92fA1",
  "publisher": {
    "id": "forbes",
    "domain": "forbes.com"
  },
  "platform": {
    "id": "openai_chat"
  },
  "citation": {
    "source_url": "https://forbes.com/crm-guide",
    "chunk_ids": ["c1"],
    "display_surface": "answer"
  }
}
```

***

## What this schema intentionally avoids

* impressions
* clicks
* dwell time
* CPM / CPX
* advertiser data
* user identifiers
* engagement scoring

Those belong to **separate systems**, not AIP v1.0.

***

## Why this design is correct

* Events are **facts**, not optimization signals.
* Publishers gain verifiable usage records.
* Platforms gain compliance proof.
* Operators can build settlement without contaminating the protocol.
* Regulators can audit end-to-end.

***

## Takeaway

If `RetrieveRequest` is **permission**, and `RetrieveResponse` is **content**, then **Events are the receipt**. You now have a complete, closed loop.

***

## Schema file

Full JSON Schema (Draft 2020-12): [access-citation-event.json](/schemas-json/access-citation-event.json)

## See also

* [RetrieveRequest](/schemas/retrieve-request)  -  Request that triggers retrieval and events
* [RetrieveResponse](/schemas/retrieve-response)  -  Response that leads to access events
* [aip.json](/schemas/aip)  -  `endpoints.event` and `required` for citation emission