YouTube Trending Scraper

Give it a list of topics. For each one it asks YouTube for that topic's videos sorted by view count, and returns the top of that list: title, channel, views, publish time, duration and thumbnail. There's no API key to get and no Google account involved, and it doesn't drive a browser.

Read this before you run it

YouTube retired the standalone Trending feed, so "what's trending" is no longer a thing you can just fetch. What this does instead is search each keyword with YouTube's own sort-by-view-count option and take the top results.

That sort is all-time, not this week. Search ai and the first rows can include an eight-year-old video with 260 million views sitting above anything from this month. That's the shape of the data, so plan around it. Two ways to do that:

• Read publishedTimeText on every row and filter yourself. It's a relative string like "3 weeks ago".
• Put the recency into the keyword. ai news august 2026 behaves very differently from ai.

Out of the box it answers "what are the biggest videos on this topic". For "what broke out this week" you need a schedule: run the same keywords daily and diff the results. Anything new in the top list is a breakout.

What you get back

{
  "ok": true,
  "videoId": "t1Pdk4VxZsw",
  "title": "Ai Ai Chand Mama and much more | Bengali Rhymes Collection",
  "url": "https://www.youtube.com/watch?v=t1Pdk4VxZsw",
  "channelName": "Infobells Bangla",
  "channelId": "UCAWuV9ZZ0itI3WR-JHWIUNQ",
  "viewCountText": "262,881,548 views",
  "publishedTimeText": "8 years ago",
  "lengthText": "19:45",
  "thumbnail": "https://i.ytimg.com/vi/t1Pdk4VxZsw/hqdefault.jpg",
  "mode": "trending"
}

View counts are text, not numbers. viewCountText is exactly the string YouTube renders — "262,881,548 views". Strip the commas and the word yourself if you need to sort on it. Same for lengthText ("19:45") and publishedTimeText ("8 years ago").

Live streams come back looking different: viewCountText reads "7,170 watching", and both publishedTimeText and lengthText are empty strings. A search on a news-heavy topic will pick up a few of those.

channelId is filled in when YouTube's payload carries it and is an empty string when it doesn't.

Input

{
  "keywords": ["ai agents", "home espresso", "padel"],
  "maxItems": 150,
  "proxyConfiguration": { "useApifyProxy": true }
}

Field	Default	Notes
keywords	empty	Required. Topics to search, one search per entry. The Console prefills music so the form runs on first click; an API call has to send its own list.
maxItems	50	Total videos across all keywords, 1 to 2,000. This is your budget cap.
proxyConfiguration	Apify proxy on	Datacenter addresses are fine for these endpoints.
notionConnector / notionParentId	—	Optional Notion delivery.

maxItems is a total, not per keyword. Three keywords and maxItems: 150 does not mean 50 each. The first keyword takes as much as it can, then the next one gets what's left. If you want an even split, run them separately.

Limits

• Videos are deduplicated on videoId across the whole run, so overlapping keywords never bill you twice for the same video.
• If one keyword fails or returns nothing, you get a diagnostic row for that keyword and the run continues with the rest. Diagnostic rows aren't billed.
• No description, no like count, no comment count. This is the search-results payload, which carries what you see on a search page and nothing more.

Billing

$0.0004 per video, which is $0.40 per 1,000, plus a $0.001 run-start fee. Flat on every Apify plan.

You're charged per video actually delivered. A run where every keyword comes back empty costs the start fee and nothing else.

Delivery

Dataset export as JSON, CSV or Excel, or read it over the API. notionConnector plus notionParentId writes one Notion page per video when the run finishes; leave them empty to skip.

What it's useful for

Sizing a niche before committing to it. You get the ceiling on view counts for a topic and the names of the channels holding those slots, which is a faster answer than scrolling YouTube for an afternoon.

channelName and channelId across a few related keywords gives you a ranked list of who's already in a space. Dedupe on channelId and count.

Thumbnail research is the third. Pull fifty rows for a topic, open the thumbnail URLs in a grid, and the house style of a niche is hard to miss.