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

# OCR and NLP response formats

> Understand the different extraction result shapes returned for OCR and NLP.

The [Extract OCR and NLP data](/api-reference/customer-v1/fax-ocr-and-nlp/post-ai-extract.mdx) endpoint uses the same response envelope for both processing types:

```json theme={null}
{
  "status": 1,
  "message": "<string>",
  "data": {
    "json_text": "<OCR array or NLP object>"
  }
}
```

The structure of `data.json_text` depends on the case-sensitive `type` supplied in the request.

## OCR results

For `type: "OCR"`, `json_text` is an array of detected document blocks. Blocks can represent a page, line, word, or another detected element. Properties vary by `BlockType`.

```json theme={null}
{
  "status": 1,
  "message": "JSON OCR data retrieved successfully.",
  "data": {
    "json_text": [
      {
        "BlockType": "PAGE",
        "Id": "0faea376-249c-4c4f-bb6f-a70241af2d65",
        "Page": 1,
        "Geometry": {
          "BoundingBox": {
            "Width": 1,
            "Height": 1,
            "Left": 0,
            "Top": 0
          }
        }
      },
      {
        "BlockType": "LINE",
        "Confidence": 100,
        "Text": "FAX",
        "Id": "27f2be95-f9d5-494b-80bd-643d75d10bc3",
        "Page": 1
      }
    ]
  }
}
```

Common OCR block properties include:

* `BlockType`: The detected element type, such as `PAGE`, `LINE`, or `WORD`.
* `Text`: Recognized text for textual blocks.
* `Confidence`: Recognition confidence returned by the OCR service.
* `Geometry`: Normalized bounding-box and polygon coordinates.
* `Id`: Unique identifier for the block.
* `Relationships`: Links to related blocks, such as a page's child elements.
* `Page`: One-based page number containing the block.

## NLP results

For `type: "NLP"`, `json_text` is an object containing processing metadata and structured values extracted from the fax.

```json theme={null}
{
  "status": 1,
  "message": "JSON NLP data retrieved successfully.",
  "data": {
    "json_text": {
      "task_id": "8cd23e57-ed34-49fd-a81c-ca068a263076",
      "action": "combined",
      "status": "success",
      "DPI": "200",
      "result_type": "object",
      "result": {
        "document_type": {
          "overall": {
            "class": "Other",
            "confidence": 0,
            "index": -1
          }
        }
      },
      "errors": []
    }
  }
}
```

The `result` object can contain document classification, patient information, sender and recipient details, insurance information, metadata, and other document-specific fields. Fields without a detected value may be empty or `null`.

<Note>
  The examples on this page are intentionally shortened. Use the response schema to integrate defensively because extracted fields depend on the document and processing type.
</Note>