AI Extract

Fetch a page and pull the fields you describe out of it. No selectors: an extraction model reads the page and returns JSON in data.json. Describe what you want with a schema (the exact shape you need), a prompt (what to look for, in words), or both; at least one is required. Every field is nullable: when the page does not contain a value, the model returns null instead of guessing. It is a good fit for one-off pages or layouts you have not mapped yet.

HTTP Request

1

POST /ai-extract

Both POST (JSON body) and GET (query-string parameters) work, run the same handler, and bill the same. POST is recommended: schema payloads are JSON objects and URLs should stay short. On GET, send schema as a URL-encoded JSON object string.

Parameters

* At least one of schema or prompt is required.

Schema shorthand

schema accepts a full JSON Schema, or a compact shorthand:

1

{"products": [{"title": "string", "price": "number", "url": "url"}]}

• Leaf types: string, number, integer, boolean, url.
• Lists: [{...}] with exactly one element describing the item shape.
• Every field is nullable: the model returns null when the page does not have it, never an invented value.
• Caps: 50 leaf fields total, nesting depth 3.

Use schema when you can: it makes the response validatable. prompt alone works too, but the model picks the JSON shape and there is no per-field validation.

Credits

Fetch credits (the same proxy_type by render_used table as /scrape) plus 5 extraction credits when json is produced:

The extraction surcharge is billed only when output was actually produced. A blocked target costs 0 credits and no extraction is attempted.

Response

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23

{
  "status": true,
  "request_id": "a35488c9-6d45-3b35-e7a4-f9e75acd",
  "data": {
    "url": "https://shop.example/category/laptops",
    "http_status": 200,
    "json": {
      "products": [
        {"title": "14-inch Ultrabook", "price": 899.0, "url": "https://shop.example/p/14-ultrabook"},
        {"title": "16-inch Workstation", "price": 1899.0, "url": "https://shop.example/p/16-workstation"}
      ]
    },
    "ai_error": null,
    "blocked": false,
    "block_reason": null,
    "detected_protection": null,
    "render_used": "http",
    "credits": 6,
    "elapsed_ms": 2140,
    "session_id": "5e4cdb8ab2a90424",
    "proxy_type": "datacenter"
  }
}

Response Fields

Error Responses

Extraction failures never turn into an HTTP error. You always get a 200 with the fetch metadata, status: false, a top-level error, and data.ai_error set; the fetch is billed, the extraction is not:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13

{
  "status": false,
  "request_id": "a35488c9-6d45-3b35-e7a4-f9e75acd",
  "error": "invalid_output",
  "data": {
    "url": "https://shop.example/category/laptops",
    "http_status": 200,
    "json": null,
    "ai_error": "invalid_output",
    "blocked": false,
    "credits": 1
  }
}

HTTP error statuses (not billed):

Code Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

import requests

url = "https://ai-web-scraper-api1.p.rapidapi.com/ai-extract"

payload = {
    "url": "https://shop.example/category/laptops",
    "schema": {
        "products": [
            {"title": "string", "price": "number", "url": "url"}
        ]
    }
}

headers = {
    "X-RapidAPI-Key": "YOUR_API_KEY",
    "X-RapidAPI-Host": "ai-web-scraper-api1.p.rapidapi.com",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)
data = response.json()["data"]

if data["json"]:
    for product in data["json"]["products"]:
        print(f"{product['title']}: {product['price']}")
        print(f"   {product['url']}\n")
else:
    print(f"Extraction failed: {data['ai_error']}")

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27

const url = "https://ai-web-scraper-api1.p.rapidapi.com/ai-extract";

const response = await fetch(url, {
  method: "POST",
  headers: {
    "X-RapidAPI-Key": "YOUR_API_KEY",
    "X-RapidAPI-Host": "ai-web-scraper-api1.p.rapidapi.com",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    url: "https://shop.example/category/laptops",
    schema: {
      products: [{ title: "string", price: "number", url: "url" }],
    },
  }),
});

const { data } = await response.json();

if (data.json) {
  data.json.products.forEach((product) => {
    console.log(`${product.title}: ${product.price}`);
    console.log(`   ${product.url}\n`);
  });
} else {
  console.log(`Extraction failed: ${data.ai_error}`);
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

curl -X POST "https://ai-web-scraper-api1.p.rapidapi.com/ai-extract" \
  -H "X-RapidAPI-Key: YOUR_API_KEY" \
  -H "X-RapidAPI-Host: ai-web-scraper-api1.p.rapidapi.com" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://shop.example/category/laptops",
    "schema": {
      "products": [
        {"title": "string", "price": "number", "url": "url"}
      ]
    }
  }'

Related Endpoints

• Generate Extraction Rules — Same input, but writes a reusable extract_rules object from sample pages; the cheap path at volume (fetch credits only on every reuse)
• Scrape — Fetch a URL and get its content plus optional CSS/XPath extraction
• AI Summarize — Fetch a URL and get a text summary instead of structured JSON