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

# Label AI-generated content

Kore.ai flags AI-generated responses so you can show AI labels in client applications and track AI messages through the Conversation History API. For channels that don't support flags, add a disclaimer to the message text or speak it in voice interactions.

***

## Platform architecture and AI detection

### Response-level flagging

* **Automatic detection:** Every response the agent sends carries a metadata flag that shows whether AI generated it.
* **Trigger condition:** The platform sets the flag when an AI model contributes to the text the end user sees.
* **Mixed content:** If a response mixes scripted static content and AI-generated content, the platform flags the whole response as AI-generated.

### Technical metadata flags

* **Primary signal:** The main flag that triggers notices is `disclaimerRequired`, sent with `isLlmGenerated`.
* **Supported delivery channels:** The platform sends these signals through the Web SDK, REST Chat API, and HTTP-async webhook payloads.
* **Audit trail:** The platform stores the `disclaimerRequired` flag with every message. Use the Conversation History API to access it for compliance audits.

### Sample response

This example shows an AI-generated response with the `isLlmGenerated` and `disclaimerRequired` flags set. The client application reads them to decide whether to show a disclosure.

```json highlight={8,12} theme={null}
{
  "sessionId": "116388f7-ab3b-4e09-9bed-5eb705c95a4b",
  "response": "Hello! Welcome to our banking service. May I have your customer ID to assist you further?",
  "action": {
    "type": "continue"
  },
  "responseMetadata": {
    "isLlmGenerated": true,
    "responseProvenance": {
      "schemaVersion": 1,
      "kind": "llm",
      "disclaimerRequired": true,
      "usedLlmInternally": true
    }
  },
  "outcome": {
    "status": "ok",
    "usedFallback": false
  },
  "sessionEnded": false
}
```

***

## Use the flag in your application

### Signal and presentation are separate

* **Platform signal:** The platform provides the technical signal (`disclaimerRequired = true`).
* **Customer UI control:** Your client application reads the flag and decides how and where to show the notice in its user interface.

### Recommended presentation options

Show the disclosure in one of these ways, based on your branding and legal requirements:

* A text label under the AI-generated message (for example, "Generated by AI").
* A visible icon or badge next to the response bubble.
* A footnote or tooltip in the response.
* A one-time notice or modal at the start of the chat session.

### Localization and customization

Your client application renders the disclosure, so you control its wording, position, and language. This lets you meet local legal and corporate requirements.

***

## Key limitations

### No built-in platform UI disclaimer

The platform has no built-in disclaimer badge and no global admin setting for disclaimer text. Build your own display logic in your frontend application or Web SDK wrapper.

### Channel-specific restrictions

Third-party messaging channels don't support custom metadata fields such as `disclaimerRequired`.

* **Affected channels:** WhatsApp, MS Teams, Slack, SMS, Email, Genesys, and Voice.
* **Workaround:** On these channels, API payload flags can't carry the disclosure. Add the disclaimer to the message body, or say it aloud in voice interactions.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.