ChatGPT Scraper API

Send prompts to ChatGPT and reliably extract the response - including the full answer text, any web-search citations, and structured metadata - at scale, without blocks.

Powered by the NetNut LLM Scraper API. NetNut handles the proxy rotation, headless rendering, anti-bot handling, and throttling, so you just send a prompt and get back clean, structured JSON.

---

Table of contents

• How it works
• Who is this for?
• Getting started
• Request sample
• Request parameters
• Output sample
• Response structure
• Code examples
• Testing
• Use cases
• Why choose NetNut
• FAQ
• Learn more

---

How it works

1. You send a POST request to the NetNut LLM Scraper API endpoint with a prompt and the chatgpt engine.
2. NetNut runs the prompt against ChatGPT using its residential proxy network and rendering infrastructure.
3. You receive structured JSON containing the ChatGPT response text and any citations from its web-search results. No HTML parsing required.

Endpoint

POST https://llm-scraper.netnut.io/search

Authentication - HTTP Basic Auth using your NetNut Scraper API credentials:

Authorization: Basic <base64(username:password)>

---

Who is this for?

This API is designed for SEO platforms, AI visibility and GEO tools, agencies, competitive intelligence providers, and enterprises that need to track, analyze, and monitor ChatGPT answers at scale.

---

Getting started

1. Get NetNut LLM Scraper API credentials. Start here or contact NetNut.

2. Copy the example environment file and add your own credentials:

   cp .env.example .env
   # then edit .env and set NETNUT_USERNAME / NETNUT_PASSWORD

   🔒 Never commit credentials. .env is git-ignored. All examples read credentials from the NETNUT_USERNAME and NETNUT_PASSWORD environment variables, nothing is hardcoded.

3. Pick your language from examples/ and run it (see each example's notes below).

---

Request sample

import os, requests

payload = {
    "prompt": "What are the top SEO trends in 2026?",
    "engine": "chatgpt",
    "web_search": True,
    "country": "us",
}

response = requests.post(
    "https://llm-scraper.netnut.io/search",
    auth=(os.environ["NETNUT_USERNAME"], os.environ["NETNUT_PASSWORD"]),
    json=payload,
    timeout=180,
)

print(response.json())

---

Request parameters

Parameter	Type	Required	Default	Description
prompt	string	Yes	-	The prompt/question to send to ChatGPT. Max length: 4096 characters.
engine	string	No	chatgpt	The chatbot engine to use. Currently only chatgpt is supported.
web_search	boolean	No	false	Enables ChatGPT's built-in web search so the answer can cite live sources.
country	string	No	-	Target country for geo-relevant results (e.g. us, uk, de).
follow_up_prompt	string	No	-	A second message sent immediately after the first response, in the same session.
raw_response	boolean	No	false	Returns the raw output from the LLM session in addition to the parsed response.

---

Output sample

{
  "timestamp": "2026-06-18T14:28:37.383Z",
  "traceID": "ffc880ee-81b1-4d55-904b-22fda6850607",
  "request_duration": 32.98,
  "process_duration": 32.98,
  "scraper": "ChatGPT",
  "chatgptAttempts": 1,
  "chromeUnblockerAttempts": 0,
  "llm_model": "gpt-5.5",
  "response": {
    "prompt": "What are the top SEO trends in 2026?",
    "text_markdown": "The biggest SEO story in 2026 is that SEO is no longer just about ranking web pages. It's increasingly about being cited, summarized, and trusted by AI-powered search systems ...",
    "text": "The biggest SEO story in 2026 is that SEO is no longer just about ranking web pages ...",
    "model_search_queries": [
      "top SEO trends 2026 AI search optimization GEO zero click search 2026",
      "SEO trends 2026 AI Overviews Google SEO 2026 predictions"
    ],
    "widgets": [],
    "citations_found": true,
    "citations": [
      {
        "id": "1",
        "title": "2026 SEO Predictions: AI, LLMs & the Future of Search",
        "url": "https://www.searcheseverywhere.com/blog/seo/2026-seo-predictions-ai-llms-future-of-search",
        "section": "citations",
        "attribution": "searcheseverywhere.com"
      }
    ],
    "tables": []
  }
}

The above is a trimmed, real response. The text / text_markdown fields contain the full answer. When web_search is enabled, citations lists the sources ChatGPT consulted (the section field is citations for sources cited inline and more for additional related sources).

---

Response structure

Field	Type	Description
timestamp	string	ISO 8601 timestamp of the response.
traceID	string	Unique identifier for the request (useful for support/debugging).
request_duration	number	Total request time in seconds.
process_duration	number	Processing time in seconds.
scraper	string	The scraper used - ChatGPT.
chatgptAttempts	number	How many attempts it took to get the answer from ChatGPT.
chromeUnblockerAttempts	number	How many unblocker attempts were needed during rendering.
llm_model	string	The ChatGPT model that produced the answer (e.g. gpt-5.5).
response.prompt	string	The prompt that was sent.
response.text	string	The full ChatGPT response as plain text.
response.text_markdown	string	The full ChatGPT response in Markdown.
response.model_search_queries	array	The web-search queries ChatGPT ran (when web_search is enabled).
response.widgets	array	Structured widgets returned in the answer (empty if none).
response.citations_found	boolean	Whether ChatGPT attached citations to the answer.
response.citations	array	List of citation objects (see below).
response.tables	array	Comparison tables extracted from the answer (empty if none).

Citation object

Field	Type	Description
id	string	Citation index (returned as a string, e.g. "1").
title	string	Source title.
url	string	Source URL.
section	string	citations for inline-cited sources, more for related sources.
attribution	string	Short domain/source attribution (e.g. searcheseverywhere.com).
description	string	Short description of the source (present on some citations).

---

Code examples

Ready-to-run examples live in examples/. Every example reads credentials from environment variables and prints the structured JSON response.

Language	Path	Notes
Python	examples/python	Uses requests.
Node.js	examples/nodejs	Uses built-in fetch (Node 18+).
Go	examples/go	Standard library only.
PHP	examples/php	Uses cURL extension.
Java	examples/java	Uses java.net.http (Java 11+).
C#	examples/csharp	.NET 8+.
Shell	examples/shell	curl one-liner.

Set your credentials first:

# macOS / Linux
export NETNUT_USERNAME="your-username"
export NETNUT_PASSWORD="your-password"

# Windows PowerShell
$env:NETNUT_USERNAME = "your-username"
$env:NETNUT_PASSWORD = "your-password"

---

Testing

Each language example ships with a unit test file. No live API calls are made during testing.

Language	Test file	Run command
Python	examples/python/test_main.py	python -m pytest test_main.py -v
Node.js	examples/nodejs/test.js	npm test or node --test test.js
Go	examples/go/main_test.go	go test ./...
PHP	examples/php/test_main.php	php test_main.php
Java	examples/java/MainTest.java	java MainTest.java
Shell	examples/shell/test.sh	bash test.sh

---

Use cases

• AI search visibility (GEO/AEO). Track whether and how your brand appears in ChatGPT answers.
• Citation/source monitoring. See which domains ChatGPT cites for the prompts that matter to you.
• Competitive intelligence. Compare how ChatGPT describes you vs. competitors across countries.
• Content optimization. Learn what ChatGPT surfaces so you can optimize content to be cited.
• Datasets & research. Collect ChatGPT responses at scale for analysis or model evaluation.

---

Why choose NetNut

• Reliable extraction at scale - proxy rotation, rendering, and anti-bot handling are built in.
• Structured JSON output - answer text + citations, no HTML scraping.
• Web search support - capture the live sources ChatGPT cites.
• Geo-targeting - get country-specific ChatGPT results.
• Backed by a large residential proxy network.

---

FAQ

What does this API do? It sends your prompt to ChatGPT and returns the response as structured JSON, including the answer text and any citations from ChatGPT's web search.

Do I need my own credentials? Yes. This repository contains no credentials. Bring your own NetNut LLM Scraper API credentials and set them as environment variables.

Is there a prompt length limit? Yes, prompts can be up to 4096 characters.

Can I enable web search? Yes, pass web_search: true to let ChatGPT search the web and cite live sources.

Can I target a specific country? Yes, pass country (e.g. us, uk, de).

What format is returned? Always structured JSON, never HTML.

---

Learn more

• 📚 NetNut LLM Scraper - ChatGPT docs
• 🌐 NetNut ChatGPT scraper

Contact us

Questions or need higher volume? Contact NetNut.

---

This project is licensed under the MIT License.