# Features
The ChatGPT Scraper provides powerful customization options to help you generate more accurate, contextual, and verifiable responses from ChatGPT. These features allow you to tailor how ChatGPT behaves, what type of responses it returns, and how much detail it extracts.
***
### Geolocation
Control and influence ChatGPT’s responses based on a target geographic region.\
Geolocation parameters affect examples, perspectives, terminology, and localized insights within the generated answer.
Geolocation supports **country-level** and **state / city-level** targeting.
***
#### Country
Influence ChatGPT’s responses by specifying a target country.
This can affect examples, perspectives, or localized insights within the answer.
**Country format**\
The `country` value must be a **2-letter country code**.
**Example**
```json
{
"prompt": "Trending products in 2025",
"country": "ca"
}
```
**Use Cases**
* Localized content generation
* Region-specific trend insights
* Country-influenced context within ChatGPT answers
***
#### State and City
Refine geolocation further by targeting a specific **state** or **state + city**.
The `state_city` parameter allows you to influence responses with more granular regional context, such as local brands, stores, events, or regulations.
> **Important**\
> The `country` parameter is **required** when using `state_city`.
**State / City format**
* State only
* State and city
* State + City must be separated by underscore
**Examples**
**State only**
```json
{
"prompt": "Trending products in 2025",
"country": "us",
"state_city": "california"
}
```
**State and city**
```json
{
"prompt": "Trending products in 2025",
"country": "us",
"state_city": "california_losangeles"
}
```
**Use Cases**
* Hyper-local trend analysis
* City-specific product or business insights
* Location-aware content and recommendations
{% hint style="info" %}
Using countries outside of the US and UK, as well as the `state_city` feature, requires a premium package.
{% endhint %}
***
### **Web Search (web\_search)**
ChatGPT includes a built-in Search mode accessible directly in its UI.\
When enabled, the scraper activates Web Search before sending your prompt.
This can lead to responses that are:
* more up-to-date
* based on external information
* more likely to include citations
#### Example
```json
{
"prompt": "Latest changes in AI regulation",
"web_search": true
}
```
#### Use Cases
* Fresh factual information
* Real-time insight generation
* Content requiring references or external validation
***
### Raw Model Output (raw\_response)
By default, the ChatGPT Scraper returns a parsed and normalized response, including text, widgets, and citations.\
For advanced use cases, you can enable the `raw_response` parameter to receive the **raw streamed output from the underlying LLM session**.
When enabled, the response will include a `raw_response` field containing the original event stream emitted by the model (for example, metadata events, delta messages, and tool outputs).
#### Parameter
* **raw\_response**: `boolean`
* **Default**: `false`
#### Example Request
```json
{
"prompt": "What are the best headphones in 2025?",
"raw_response": true
}
```
#### Example Response
{
"traceID": "71f9e0b2-3a77-4ad0-9a61-23b7e1bfa8d1",
"model": "gpt-5.2",
"timestamp": "2025-12-08T11:39:19.269Z",
"request_duration": 6.3,
"process_duration": 7.0,
"response": {
"text_markdown": "Here are some of the **best headphones you can buy in 2025**...",
"text": "Here are some of the best headphones you can buy in 2025...",
..
"raw_response": [
"data: {\"type\":\"server_ste_metadata\",\"metadata\":{\"model_slug\":\"gpt-5.2\"}}",
"event: delta",
"data: {...}"
]
}
}
#### Notes
* `raw_response` is **optional** and disabled by default to reduce payload size.
* When enabled, it does **not replace** parsed fields (`text`, `widgets`, `citations`), it is returned **in addition** to them.
***
### **Follow-up Prompts (follow\_up\_prompt)**
Send a second prompt in the same ChatGPT session.\
The follow-up prompt runs after the first answer is fully generated.
#### Example
```json
{
"prompt": "List ranking factors",
"follow_up_prompt": "Summarize them into three main themes"
}
```
#### Use Cases
* Refinement of initial answers
* Multi-step reasoning
* Structuring, summarizing, or analyzing generated content
***
### **Callback URL (callback\_url)**
\
Send the LLM Scraper response directly to your own API endpoint.\
When `callback_url` is provided, the request is acknowledged and the final LLM Scraper response is delivered to your endpoint via a POST request.
**Example Request**
```json
{
"prompt": "Summarize the main benefits of residential proxies",
"callback_url": "https://user-api.com"
}
```
**Initial Response (202 Accepted)**\
When `callback_url` is included, the API returns a confirmation response indicating that the request was received.
```json
{
"message": "request received",
"traceId": "123e4567-e89b-12d3-a456-426614174000",
"response_url": "https://user-api.com"
}
```
**Final Response**\
Once processing is completed, the full LLM Scraper response will be sent to the provided `callback_url`.\
The response structure is identical to the standard LLM Scraper response returned in synchronous requests.
**Use Cases**
* Deliver results directly into backend workflows
* Separate request handling from response processing
* Trigger automation pipelines after completion
***
### Automated Throttling (`throttling`)
The service includes an automated throttling mechanism designed to dynamically manage requests and maintain a high success rate during heavy traffic or platform-side limitations.
By default, throttling is enabled.
For advanced use cases, customers may choose to disable throttling by setting the `throttling` parameter to `false`.
#### Parameter
* type: `boolean`
* Default: `true`
#### Example Request
```json
{
"prompt": "What are the best headphones in 2025?",
"throttling": true
}
```
***
### **Response Schema Overview**
Each session includes:
* `traceID` for request tracking and debugging
* `timestamp` indicating when the request was processed
* `request_duration` and `process_duration` for performance visibility
* `model` identifying the AI model used (for example, `gpt-5.2`)
* Parsed response content in both markdown and plain text
* Structured widgets extracted from the model output (when available)
* Citation metadata with section classification
* Model-generated search queries used during retrieval (when applicable)
* Raw streamed model output for debugging and advanced use cases
* Follow-up response content if an additional prompt was provided
#### Example (condensed)
```json
{
"traceID": "71f9e0b2-3a77-4ad0-9a61-23b7e1bfa8d1",
"model": "gpt-5.2",
"timestamp": "2025-12-08T11:39:19.269Z",
"request_duration": 6.3,
"process_duration": 7.0,
"response": {
"text_markdown": "Here are some of the **best headphones you can buy in 2025**...",
"text": "Here are some of the best headphones you can buy in 2025...",
"model_search_queries": [
"best headphones 2025 reviews and ranking",
"top headphones to buy in 2025 best headphones list"
],
"widgets": [
{
"type": "product_carousel",
"items": [
{
"type": "product_card",
"name": "Sony WH-1000XM6",
"price": 399.99,
"currency": "$"
}
]
}
],
"citations_found": true,
"citations": [
{
"id": 1,
"url": "https://example.com/best-headphones-2025",
"title": "Best Headphones of 2025",
"section": "citations"
}
]
},
"raw_response": [
"data: {\"type\":\"server_ste_metadata\",\"metadata\":{\"model_slug\":\"gpt-5.2\"}}",
"event: delta",
"data: {...}"
]
}
```
***
### **Error Handling**
The ChatGPT Scraper includes standardized error responses.
#### **400 — Validation Errors**
Returned when:
* Prompt missing
* Invalid parameter
* Prompt > 4096 characters
Example:
```json
{
"message": "Prompt exceeds 4096 characters. Please shorten your prompt.",
"error": "Bad Request",
"statusCode": 400
}
```
***
#### **401 — Credentials Error**
Returned when:
* Username or Password is Incorrect
* No Active Package for the Product
Example:
```json
{
"message": "Request failed with status code 401"
}
```
***
#### **500 — Internal Errors**
Every Error of a Scraping error will return a 500 error code like so:
```json
{
"message": "The ChatGPT scraper could not complete the request. Please try again.",
"error": "Internal Scraper Error",
"statusCode": 500
}
```
---
# Agent Instructions: Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter:
```
GET https://help.netnut.io/netnut-documentation/netnut-scraper-apis/llm-scraper/chatgpt/features.md?ask=
```
The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.