Discover
Test a source before building a dedicated scraper.
Extract website data with AI without writing selectors
Describe the columns you need in natural language and let the Actor interpret each page, useful for prototypes where structure varies or CSS is unknown.
The first button is an affiliate link. We may receive a commission if you purchase an eligible plan, at no extra cost to you.
Direct answer
How do you scrape a website with AI and a prompt?
Provide one or more URLs, choose single to process them individually and write a prompt with fields, types, missing-value rules and output format. Limit pages and depth before trying scout or agentic modes.
AI reduces setup, not errors: compare a sample against the original page and measure field-level accuracy.
- Actor
apify/ai-web-scraper- Input
- URL, query and scope limits
- Output
- Structured JSON defined by the prompt
- Best for
- Extraction prototypes and variable pages
- Checked
- July 24, 2026
Actor used
AI Web Scraper
Built by Apify. Extraction prototypes and variable pages.
Check the official listing4.26 / 512 reviews
8.1Ktotal users
283monthly users
154bookmarks
Public listing signals observed July 24, 2026; they can change.
Copy-ready setup
Small input for validation
{
"startUrls": [{"url": "https://example.com/pricing"}],
"extractionMode": "single",
"prompt": "Return product_name as text, monthly_price as number, currency as ISO code, and features as an array. Use null when absent.",
"maxPagesToVisit": 10,
"maxCrawlDepth": 2
}- 01
Define a small schema
Start with four or five fields you can verify.
- 02
Specify types and missing values
Specify numbers, arrays, dates and null to avoid ambiguous output.
- 03
Validate page by page
Measure completeness and accuracy against the source.
- 04
Scale the mode carefully
Use scout or agentic only when single cannot handle navigation.
Fields to review
What to check before using the Dataset
| Field | Use and check |
|---|---|
prompt | extraction contract |
extractionMode | control autonomy |
maxPagesToVisit | cap cost and scope |
Dataset JSON | inspect types and nulls |
Cost
Estimate the main Free event
Each run start costs $0.00025; complex pages may require more navigation.
Free estimate$3.00
Main event before add-ons and plan credits.
| Public rates by tier | USD |
|---|---|
| Free | $30.00 per 1,000 pages |
| Bronze | $25.00 per 1,000 |
| Gold | $20.00 per 1,000 |
Each run start costs $0.00025; complex pages may require more navigation.
Added value
Why this Actor can justify trying Apify
Compare
Normalize pages with different layouts.
Automate
Create small datasets for research or automation.
FAQ
Before you scale
Do I need to code?
Not for the basic input, but you must define and validate the data schema.
Is it more accurate than selector-based scraping?
It depends on the page. It is more flexible, while deterministic extraction is often more stable at scale.
Which mode should I try first?
single, because it limits navigation and makes accuracy and cost easier to measure.
First run
Validate a small sample before automating
Confirm relevance, fields and cost with the input above.