YouTube Video Scraper — Views, Likes & Metadata

Scrape any YouTube video without an API key, a quota, or a login: title, channel, view count, like count, duration, publish date, category, keywords and thumbnail.

Paste watch URLs, youtu.be links, Shorts, embeds or bare video IDs. All forms work and can be mixed in one run.

Output

{
  "videoId": "dQw4w9WgXcQ",
  "url": "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
  "title": "Rick Astley - Never Gonna Give You Up (Official Video) (4K Remaster)",
  "channelName": "Rick Astley",
  "channelId": "UCuAXFkgsw1L7xaCfnd5JJOw",
  "channelUrl": "https://www.youtube.com/channel/UCuAXFkgsw1L7xaCfnd5JJOw",
  "viewCount": 1815622496,
  "likeCount": 19390234,
  "lengthSeconds": 213,
  "publishDate": "2009-10-24T23:57:33-07:00",
  "category": "Music",
  "keywords": ["rick astley", "nggyu"],
  "isLive": false,
  "unavailable": false,
  "availableCaptionLanguages": ["es-419", "de-DE", "en"],
  "thumbnail": "https://i.ytimg.com/...",
  "description": "..."
}

Three things this gets right

Like counts still work. Likes left the YouTube watch page during 2026 — they are in neither embedded blob. Scrapers that never noticed now return null or a stale guess. This Actor reads them from YouTube's own internal endpoint, verified live: 19,390,234 on the video above.

Because that costs a second request per video, includeLikes is a switch. Turn it off and each video costs one request instead of two.

An unavailable video is an answer, not a failure. Private, deleted and region-blocked videos are reported with unavailable carrying YouTube's own reason, and the run continues. The row keeps the same field set as every other row, so exports stay rectangular.

The parser reports its own breakage. Required fields are checked after extraction. If YouTube changes the page and a title or view count stops arriving, the run logs it and RUN_SUMMARY counts it, instead of quietly emitting a dataset full of nulls. A live stream with no view count is explicitly not treated as breakage.

Input

Field	What it does
videos	URLs or 11-character IDs. Watch, youtu.be, Shorts, embed and live all accepted.
includeLikes	Fetch like counts. Costs one extra request per video. On by default.
includeDescription	Turn off for a compact dataset.
language / country	Two-letter codes. Country affects availability and regional blocking.
maxItems	Hard cap on videos processed, and therefore on cost.

Example

{
  "videos": [
    "https://www.youtube.com/watch?v=dQw4w9WgXcQ",
    "https://youtu.be/9bZkp7q19f0",
    "jNQXAC9IVRw"
  ],
  "includeLikes": true
}

Proxies and reliability

YouTube intermittently answers datacenter IPs with "Sign in to confirm you're not a bot". Measured on Apify infrastructure on 14 September 2026: some videos returned normally, others were refused, with and without datacenter proxy. Routing through residential IPs returned every video successfully.

Two things follow, and both are handled rather than hidden:

A blocked video is never charged. A bot check means this Actor's IP was refused. The row is still written to the dataset, flagged blocked: true with error: "blocked_by_youtube", so you can see exactly what happened, but no event fires for it. You pay for videos you actually receive.

A blocked video is not confused with a deleted one. A private, removed or region-blocked video is a real answer and is reported through unavailable with YouTube's own wording. Only the bot check sets blocked.

To collect every video reliably, set proxyConfiguration to Apify Proxy with RESIDENTIAL groups. Note that residential bandwidth is billed at $8/GB and a watch page plus the likes call is roughly 2.4 MB per video, so residential runs carry a real cost. For metadata-only runs, switch includeLikes off to roughly halve the bandwidth.

Notes and limits

• Public video data only. No login, no cookies, no Google account.
• Transcripts are not included. Caption languages are reported, but the caption tracks themselves now require a proof-of-origin token and return empty without one. Rather than ship a transcript feature that silently returns nothing, it is left out.
• Comment counts are not returned for the same reason: they are no longer on the page. Comments are a separate job needing continuation pagination.
• YouTube throttles bursts with HTTP 429. Requests are spaced and retried with exponential backoff, honouring Retry-After.

Development

npm install
npm test
node src/main.js