Scrape YouTube video data into clean JSON

One call

Send the URL and a plain-language description of the fields. That's the whole integration.

curl https://api.webscrape.ai/v1/smartscraper \
  -H "X-API-Key: wsg_live_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "website_url": "https://www.youtube.com/watch?v=jNQXAC9IVRw",
    "user_prompt": "Extract the video title, channel name, view count, like count, and publish date."
  }'

You get back JSON shaped to match — nothing to parse:

{
  "status": "completed",
  "data": {
    "result": {
      "title": "string",
      "channel": "string",
      "views": "number",
      "likes": "number",
      "published": "string"
    }
  },
  "credits_used": 5,
  "credits_remaining": 495,
  "request_id": "req_a1b2c3d4"
}

Example response shape; data.result matches the fields you asked for. Failed fetches return a non-completed status and cost nothing.

Fields you'll usually want from YouTube

YouTube exposes an official Data API too; the page route is handy for quick public metadata. Respect YouTube’s terms.

How it works

• Tiered fetching. A fast request carrying a real browser's fingerprint handles most pages. When a site needs a full browser, turn on stealth mode and the request runs through a hardened stealth browser instead.
• AI extraction. The cleaned page is extracted into data matching your fields; pass an output_schema and the result is validated against it, with one repair pass on a miss.
• One wallet, success-only. Proxies, browser time, and extraction all come out of one credit balance, and a failed fetch costs nothing.
• Blueprints for repeat runs. Once an extraction works, pin it: subsequent runs replay reverse-engineered selectors with no AI step in the loop.

When it breaks

• A page returns a 200 with an empty shell: that can look like success to the fast tier, so force the browser tier for that request if it recurs.
• A field comes back wrong after a redesign: re-run live extraction, or re-pin the Blueprint. Drift surfaces as validation_warnings, not a silent null.

FAQ