Bilibili Video Search Scraper

Search the biggest video community in China by keyword and get structured results. Each row is a video with its cover image, title, creator, views, likes, comments, danmaku (on-screen comment) count, favorites, duration, tags and publish time.

No login, no cookies, no browser needed. Results are saved as a dataset you can download as JSON, CSV or Excel, or read through the Apify API. The first column of the table is the thumbnail image.

What you get

• Sort by relevance, newest, most viewed, most danmaku or most favorited.
• Up to 1,000 videos per keyword (Bilibili lists 50 pages of 20).
• Clean titles with the search highlighting removed, plus tags, category and the creator profile link.
• Every field Bilibili returns is kept; anything without its own column is under "Other data".

Who uses this actor

• Trend and topic research on Chinese-language video.
• Find creators and videos in a niche for outreach.
• Track how a keyword performs over time.
• Competitor and brand monitoring.
• Datasets for content analysis.

How to use it

1. Open the Input tab. The fields already hold a working example.
2. Click Start. Small runs finish in seconds.
3. Open the Output tab to see the table, then export it or connect it to your tools.

If something cannot be read, for example an unknown user or a video that does not exist, the run still finishes successfully and says what was wrong in the run status message, with an example of a working input. Results are saved while the run is going, so you can start using them straight away.

Input

Field	Name	Required	Default	What it does
keyword	Keyword	yes	example: "python"	What to search for on Bilibili. Chinese or English, for example "python" or "原神".
order	Sort by	no	"relevance"	How Bilibili should order the results.
maxItems	Maximum videos	no	40	Stop after this many unique videos. Bilibili lists at most 1,000 per keyword, so you may get fewer.

Example input

{
  "keyword": "python",
  "order": "newest",
  "maxItems": 40
}

Output

Each record is one row in the dataset. Fields Bilibili does not provide for a record are left out, and numbers that are really zero stay as 0. Field names are plain English so tables read well in Excel and CSV.

Field	Type	Meaning
Cover image (image)	string	Video cover (shown as a thumbnail).
Video ID (id)	string	Bilibili BV id, unique per video.
Title (title)	string	Video title, plain text.
Video link (url)	string	Link to the video on Bilibili.
Description (description)	string	Video description text.
Creator (creator)	string	Uploader name.
Creator ID (creatorId)	string	Uploader user id (mid).
Creator link (creatorUrl)	string	Link to the uploader space page.
Creator avatar (creatorAvatar)	string	Uploader profile picture.
Category (category)	string	Bilibili category of the video.
Duration (seconds) (durationSeconds)	number	Video length in seconds.
Published at (publishedAt)	string	Publish time, ISO 8601 (UTC).
Views (views)	number	Play count.
Likes (likes)	number	Like count.
Danmaku (danmaku)	number	Number of danmaku (on-screen comments).
Comments (comments)	number	Number of comments.
Favorites (favorites)	number	How many users saved the video.
Tags (tags)	array	Tags the creator added to the video.
Matched in (matchedIn)	array	Which part of the video matched your keyword, for example title or tag.
Paid video (isPaid)	boolean	True for paid or charging videos.
Collaboration (isCollaboration)	boolean	True when the video is a joint upload by several creators.
Badge (badge)	string	Corner badge text, when Bilibili shows one.
Recommend reason (recommendReason)	string	Why Bilibili recommends the result, when given.
AV id (avId)	string	Numeric av id of the video (the older Bilibili id style).
Position (position)	number	Position of the video in the result list (1 is first).
Search keyword (searchKeyword)	string	The keyword that produced this row.
Scraped at (scrapedAt)	string	When this record was collected, ISO 8601.
Other data (otherData)	object	Any additional field Bilibili sent that has no column of its own. Empty placeholder values are left out.

Example output

{
  "image": "https://i0.hdslb.com/bfs/archive/c92ffada9fbaa96ba1e146200e5421325516ed4d.jpg",
  "id": "BV1kRpW6AEhs",
  "title": "【双语字幕】Pragmatic AI Labs:从零开始学习 WebAssembly",
  "url": "https://www.bilibili.com/video/BV1kRpW6AEhs",
  "creator": "原声译站",
  "creatorId": "3546981175658915",
  "creatorUrl": "https://space.bilibili.com/3546981175658915",
  "creatorAvatar": "https://i2.hdslb.com/bfs/face/f20137060042a192777c566b537b78455545404b.jpg",
  "category": "计算机技术",
  "durationSeconds": 2314,
  "publishedAt": "2026-10-08T21:11:53.000Z",
  "views": 5,
  "likes": 1,
  "danmaku": 0,
  "comments": 3,
  "favorites": 0,
  "tags": [
    "沙箱代码",
    "可移植性",
    "嵌入式设备",
    "WASM",
    "WebAssembly",
    "Rust",
    "JavaScript",
    "Python"
  ],
  "matchedIn": [
    "tag"
  ],
  "isPaid": false,
  "isCollaboration": false,
  "avId": "117407343511051",
  "position": 1,
  "searchKeyword": "python",
  "scrapedAt": "2026-10-08T22:31:49.686Z"
}

Tips

• Chinese keywords usually give the widest coverage. Titles and comments come back in their original language.
• The actor paces its requests to stay within Bilibili's limits, so larger runs take a little longer. Spread very large jobs over several runs or schedules.
• Export as JSON, CSV or Excel, or read the results through the Apify API.

FAQ

How many videos can I get per keyword? Bilibili lists up to 1,000 results (50 pages of 20) per keyword and sort order. Use different keywords or sort orders to reach more.

Why are views 0 for brand-new videos? Bilibili reports the play count it has at that moment; videos published minutes ago can still show 0.

Is it legal to scrape Bilibili? The actor reads only data that Bilibili shows publicly, without logging in. Comments and profiles can contain personal data, so use the results in line with the data protection rules that apply to you and with Bilibili's terms.

Something looks wrong. What should I do? Open the run log and the run status message first; they explain what happened. If the data still looks wrong, send the run link through the Issues tab and it will be looked at.