YouTube Trending Scraper

See exactly what's taking off on YouTube right now. Pick a category and a country and this actor hands you one tidy row per trending video — no YouTube Data API key, no headless browser, no login required.

Under the hood it speaks directly to YouTube's own internal InnerTube API (the very endpoints the YouTube app and website call), which keeps every run fast, cheap and dependable.

What you get

One dataset row per trending video:

Field	Description
category	Trending category (now, music, gaming, movies)
source	Which source produced the row — see How it works
rank	1-based position in the trending list
videoId	11-character video ID
title	Video title
url	https://www.youtube.com/watch?v=<videoId>
channelName	Channel display name
channelId	The channel's UC… ID (when available)
channelUrl	https://www.youtube.com/channel/<channelId> (when available)
viewsText	View count as YouTube's own display text, e.g. "15,780,087 views"
viewCount	Best-effort integer parse of viewsText, e.g. 15780087
publishedText	Relative publish time, e.g. "5 days ago"
lengthText	Duration, e.g. "3:25"
descriptionSnippet	Short description snippet (when available)
thumbnail	Highest-resolution thumbnail URL
thumbnails	Full array of { url, width, height } thumbnails

If nothing can be fetched for the requested region, the actor still writes a single row — { "category": "…", "region": "…", "error": "…" } — so a run is never silently empty.

Input

Field	Type	Required	Default	Description
category	string	no	now	now (general), music, gaming, or movies
maxResults	integer	no	50	Maximum number of trending videos to return
gl	string	no	US	ISO country code for the trending region (e.g. US, GB, DE, IN)
country	string	no	—	Alias for gl; gl wins if both are set
language	string	no	en	ISO language code for titles / metadata
proxyCountryCode	string	no	US	ISO country code for the Apify residential proxy exit

Example input

{
    "category": "now",
    "maxResults": 50,
    "gl": "US",
    "proxyCountryCode": "US"
}

Example output row

{
    "category": "now",
    "source": "most_viewed_this_week",
    "rank": 1,
    "videoId": "qFhspt7zsXg",
    "title": "Number_i - BUGS LIFE (Official Music Video)",
    "url": "https://www.youtube.com/watch?v=qFhspt7zsXg",
    "channelName": "Number_i OFFICIAL",
    "channelId": "UC7N348fUF5Sh-N4m1y578zg",
    "channelUrl": "https://www.youtube.com/channel/UC7N348fUF5Sh-N4m1y578zg",
    "viewsText": "15,780,087 views",
    "viewCount": 15780087,
    "publishedText": "5 days ago",
    "lengthText": "3:25",
    "thumbnail": "https://i.ytimg.com/vi/qFhspt7zsXg/hqdefault.jpg"
}

How it works

YouTube retired its dedicated Trending feed (browseId: "FEtrending") in 2025: every InnerTube client now returns HTTP 400 for it, /feed/trending redirects to the (empty, logged-out) home feed, and the feed no longer appears in the navigation guide. To keep returning a real, up-to-date trending list, this actor uses a resilient two-step strategy:

1. Legacy Trending feed (forward-compatibility). It first POSTs to youtubei/v1/browse with browseId: "FEtrending" (and the classic category params) using the WEB client. If YouTube ever restores this feed, the actor picks it up automatically and rows are tagged source: "trending".
2. Most-viewed-this-week search (the working source today). When the legacy feed is unavailable, it POSTs to youtubei/v1/search with the ANDROID client, sorting by view count and restricting to videos uploaded this week (search filter params: "CAMSBAgDEAE="). A broad seed query per category (now uses a ubiquitous stop-word so the result reflects the week's overall most-viewed videos across every category) surfaces what is currently blowing up. It pages through continuation tokens until maxResults is reached, de-duplicating by videoId and guarding against continuation loops. These rows are tagged source: "most_viewed_this_week".

Requests are routed through Apify residential proxy when available, with an automatic fallback to a direct connection so a blocked or absent proxy never crashes the run.

Pricing

This actor uses Apify's pay-per-event model. You're charged once per trending video delivered to the dataset (video-scraped). Error rows are never billed, and the run stops automatically the moment your budget is reached.

Notes & limitations

• Because YouTube removed the official Trending feed, "trending" is currently derived from the most-viewed videos uploaded in the last week for the chosen category and region. This closely tracks what people are watching now, but it is not identical to the old editorially-curated Trending page.
• viewsText, publishedText and lengthText are YouTube's own localized display text; viewCount is a best-effort numeric parse of viewsText.
• The region is controlled by gl / country; for the most accurate regional results, set proxyCountryCode to the same country.
• YouTube can change its internal API at any time; this actor tolerates several response shapes but is not covered by an official API contract.

Tech

Written in TypeScript on the Apify SDK, using got-scraping for the InnerTube requests. HTTP-only, so builds are fast and runs are cheap.

License

MIT © Assia Fadli