Response Fields

This page explains the data returned by the LLM Scraper. Each field is briefly described, with an example where relevant, so teams can quickly understand and integrate the response.

Top-Level Fields

Field	Description	Example
traceID	Unique identifier for the request. Used for debugging, support, and internal tracking.	"traceID": "71f9e0b2-3a77-4ad0- 9a61-23b7e1bfa8d1"
timestamp	ISO-8601 timestamp indicating when the request was processed.	"timestamp": "2025-12-08 T11:39:19.269Z"
request_duration	End-to-end request duration in seconds (from request received to response returned).	"request_duration": 6.3
process_duration	Total internal processing time in seconds. This includes the full request lifecycle.	"process_duration": 6.4
model	LLM model used to generate the response.	"model": "gpt-5.2"

Note request_duration is included within process_duration. For example, if request_duration is 6.3 and process_duration is 6.4, the overall time experienced by the client is 6.4 seconds.

---

Response Object

The main parsed output from the LLM.

---

prompt

The original prompt sent to the LLM for this request. This field is always returned as part of the response for traceability and debugging.

Example

"prompt": "what is the best seo tips in 2025"

---

text_markdown

Markdown-formatted version of the answer, preserving headings, lists, emphasis, and links.

Example

"text_markdown": "Here are the **leading SEO tips for 2025**:\n\n### 1. High-quality content\n### 2. Strong UX"

---

text (plain text)

Plain-text version of the response with formatting removed. Useful for storage, indexing, or systems that don’t support Markdown.

Example

"text": "Here are the leading SEO tips for 2025 include focusing on content quality and user experience."

---

model_search_queries

Search queries generated by the model during Web Search mode. These reflect how the model expanded or refined the original prompt.

Example

"model_search_queries": [
  "leading SEO tips in 2025",
  "top SEO best practices 2025"
]

---

widgets

Structured UI components extracted from the model output, such as product carousels or cards.

Example

"widgets": [
  {
    "type": "product_carousel",
    "items": [
      {
        "type": "product_card",
        "name": "Headphones A",
        "price": 399.99,
        "currency": "$",
        "Images": ["https://.."],
        "merchants": [
          {
            "merchant": "Merchant A",
            "price": 399.99,
            "currency": "$",
            "url": "https://merchant-a.example/product/headphones-a"
          },
          {
            "merchant": "Merchant B",
            "price": 409.99,
            "currency": "$",
            "url": "https://merchant-b.example/product/headphones-a"
          }
        ]
      }
    ]
  }
]

---

citations_found

Boolean indicating whether the model output included citations.

Example

"citations_found": true

---

citations

Flat list of citation objects extracted from the ChatGPT UI.

Each citation includes:

• id

• title

• url

• section: "citations" or "more"

Example

"citations": [
  {
    "id": 1,
    "title": "Best Headphones of 2025",
    "url": "https://example.com/best-headphones-2025",
    "section": "citations"
  }
]

---

text_html

Raw HTML representation of the model’s rendered answer, as produced by the ChatGPT UI.

This field contains the original HTML markup of the response content (including layout structure, classes, and formatting).

Example

"text_html": "<div class=\"min-h-8 text-message relative flex w-full flex-col items-end gap-2 text-start break-words whitespace-normal [.text-message+&amp;]:mt-1\" data-message-author-role=\"assistant\" data-message-id=\"3b9247fc-daf4-480c-8a8b-69aa11bb8393\" data-message-model-slug=\"gpt-5-2\" dir=\"auto\">\n <div class=\"flex w-full flex-col gap-1 empty:hidden first:pt-[1px]\">\n  <div class=\"markdown prose dark:prose-invert w-full wrap-break-word light..."

---

raw_response (Optional)

Raw streamed output from the underlying LLM session. Returned only when raw_response: true is set in the request.

Includes metadata events, delta messages, and tool outputs.

Example

"raw_response": [
  "data: {\"type\":\"server_metadata\",\"metadata\":{\"model_slug\":\"gpt-5.2\"}}",
  "event: delta",
  "data: {...}"
]

Notes

• Disabled by default

• Returned in addition to parsed fields

• Intended for advanced use cases