Weibo Scraper: Posts, Comments & Hot Search

What is Weibo Scraper?

Weibo (微博) is where Chinese public opinion moves first, and its web pages are built for scrolling, not for counting. Copying posts, comments and numbers into a spreadsheet by hand stops working after a few dozen.

This Actor reads public Weibo data without a login and gives you clean rows:

• Posts by keyword — Weibo's real-time search, newest first.
• Posts by account — an account's recent posts, plus its public profile.
• Posts by URL, with their comments — every comment with the commenter's region (来自 X), its floor number, its like count and whether the post's author liked it, plus replies if you want them.
• The hot search list (微博热搜) — the live top 50 with heat values, labels and the pinned entry.

Every post row carries what makes Weibo data useful beyond the text: video play counts, how many times the post was edited, AI-generated / reposted content labels, t.cn short links expanded to where they point, all picture sizes and video files, hashtags, mentions, the province it was posted from, and the full original when it is a repost.

Because it runs on Apify, you can schedule a daily keyword or hot-search snapshot, call it from the API, connect it to Make, Zapier or Google Sheets, and export as CSV, Excel or JSON.

How to scrape Weibo

1. Open the Input tab. Fill in any of the sections; they all work together in one run.
   • Search keywords: 瑞幸咖啡, #新能源汽车#, a brand name.
   • Accounts: 1669879400, https://weibo.com/u/1669879400 or @Dear-迪丽热巴.
   • Post URLs: https://weibo.com/1669879400/RlZog45UR or https://m.weibo.cn/detail/5352408760976057.
   • Hot search: tick the box.
2. Set how many posts and comments you want. Each row is charged, so the limits are also your budget.
3. Click Start. The prefilled example (10 posts, one post with 10 comments, the hot search list) finishes in about half a minute.
4. Open the Output tab. There is one view per row type: Posts, Comments, Profiles, Hot search, and Inputs with no rows. A view picks the columns, not the rows: rows of other types appear in it with those columns empty, so filter on type when you export.

What data does it return?

Every row has a type (post, comment, profile, trend or status), the input that produced it, and scrapedAt. Times are ISO-8601 in Beijing time (2026-10-18T11:35:55+08:00). Counts are numbers. A value Weibo does not show is null, never a made-up 0.

Posts (type: "post")

Field	Notes
mid, bid, url	Numeric ID, short ID and link
createdAt	When it was posted
text, textHtml	Plain text and Weibo's HTML
isLongText, textExpanded	Long posts are cut short in feeds; with Expand long posts on, the full text is fetched
repostsCount, commentsCount, likesCount	Exact numbers
countsCapAt, cappedCounts	Weibo stops counting publicly at 1,000,000 ("100万+"). A count at the cap is listed in cappedCounts, so you know it is a lower bound
videoPlayCount, videoPlayCountText	Video plays as a number, and Weibo's text ("10.5万次观看")
videos	Every video: title, duration, play count, cover, every quality's file URL
pics	Every picture with all sizes Weibo serves (thumbnail to mw2000), live-photo videos included
editCount, isEdited	How many times the author edited it
aiTags	Weibo's labels such as "博主声明：该内容使用AI生成技术" (AI-generated) or "内容为转载" (reposted content)
contentLabels	Declarations such as "自主创作" (original work)
links	Every link with the t.cn short link, the expanded long URL and its title
hashtags, topics, superTopic, mentions
regionName, regionRaw	"河北" from "发布于 河北": the province the post was published from
source	Device or client, e.g. "iPhone 18 Pro"
card, audio	Attached topic, live stream or web card; attached audio
isRepost, retweetedStatus	The original post, with all the same fields
isAd, isPaid, coAuthors
author	uid, screenName, profileUrl, avatarUrl, verified, verifiedType, verifiedReason, memberRank, followersCount (see below)

Feeds do not include an author's follower count, so author.followersCount is null unless you turn on Add author profiles to posts.

Comments (type: "comment")

Field	Notes
id, postMid, postUrl
level, rootId, replyToCommentId, replyToUser	level 0 is a comment, 1 a reply
createdAt, text, textHtml	Emoticons kept as [name]
likesCount, repliesCount
region, regionRaw	"福建" from "来自 福建": where the commenter posted from
floorNumber	The comment's floor (楼) under the post
isLikedByMblogAuthor	Whether the post's author liked this comment
picUrls, links	Pictures attached to the comment, links in it
previewReplies	The first replies Weibo shows under a comment, nested (free)
author	The commenter's public account fields, as for posts

Profiles (type: "profile")

uid, screenName, profileUrl, customUrl, avatarUrl, coverImageUrl, description, location, ipLocation, followersCount (+ followersCountText such as "8255.1万"), followingCount, postsCount, totalInteractions, totalLikes, totalComments, totalReposts, verified, verifiedType, verifiedReason, memberRank, isSvip, website, registeredAt, sunshineCredit, and Weibo's influence labels ("V指数 文娱 87.49分", "近30天视频播放量240.2万").

Hot search (type: "trend")

rank, position, word, note, query, url (the search page), hotValue, label (新, 热, 沸, 爆), category, isTopic, isAd, isPinned, isGovernment, capturedAt. Ads in the list are kept, marked isAd, and have no rank.

Inputs with no rows (type: "status")

Every input is accounted for. An input that produced no rows, or stopped early, gets a free row with status and a plain error:

no_results, not_found, invalid_input, duplicate_input, blocked, partial, parse_error, not_processed. The key-value store record INPUT_SUMMARY lists every input with its status and row count.

How much does it cost to scrape Weibo?

You pay per row, and nothing for time spent waiting on Weibo.

Event	Price	Per 1,000
Post (post-scraped)	$0.008	$8
Comment or reply (comment-scraped)	$0.004	$4
Profile (profile-scraped)	$0.008	$8
Hot search entry (trend-scraped)	$0.002	$2
Run start (apify-actor-start, per GB of memory)	$0.001	—

Examples: 500 keyword posts cost $4. One post with 1,000 comments costs $4.008. A daily hot-search snapshot (about 52 entries) costs about $0.10.

Status rows, posts or comments Weibo served in a broken shape (skipped and counted in the input's error), expanded long text, the preview replies nested under each comment, reposts you chose to skip and residential proxy traffic are free. A post that appears under two keywords is written, and charged, once.

How do I cap what a run can spend?

Set a maximum cost per run in the run options. Before each page it checks that the limit can still pay for a row, and rows are written and charged together, so a page that only partly fits writes and charges just the rows that fit. Inputs it did not get to come back as not_processed rows. If writing or charging fails, the run stops instead of working unpaid, and the charge is not sent a second time.

You are never charged for a row that is not in your dataset. If Apify moves the run to another server mid-way, the restarted run checks its dataset and does not write or charge a row a second time.

Input example

{
  "keywords": ["瑞幸咖啡"],
  "maxPostsPerKeyword": 100,
  "users": ["1669879400"],
  "maxPostsPerUser": 20,
  "postUrls": ["https://weibo.com/1669879400/RlZog45UR"],
  "maxCommentsPerPost": 200,
  "commentsSort": "hot",
  "includeHotSearch": true
}

Limits worth knowing

• Keyword search reaches about 1,000 posts per keyword, newest first: about 24 pages. For a busy keyword that is the last few hours. Weibo's own "hot" and date filters live on s.weibo.com, which refuses logged-out visitors, so there is no sort option; use Only posts published on or after to cut old posts.
• Account posts can have gaps. Logged out, Weibo refuses the account timeline itself, so posts come from Weibo's search index of that account. It returns about 50 posts a page and goes back years, but it can skip posts: for one account we tested, the index held 1,018 of its 1,937 posts. Use it for recent activity and trends, not as a complete archive.
• Reposts of a post (转发列表) are not available logged out; the repost count is.
• Custom profile addresses (weibo.com/name) cannot be resolved logged out. Use the numeric ID or @screen name.
• Counts of 1,000,000 or more are lower bounds, because Weibo shows "100万+". They are flagged in cappedCounts.
• The original inside a repost is not expanded when it is long: retweetedStatus.text is Weibo's preview. Pass its url under Post URLs for the full text.
• Picture and video URLs are signed and expire after a while (the Expires parameter). Download files you need soon after the run.
• Weibo throttles visitors. When it answers "前方有点拥堵", the Actor waits, takes a new IP and a new visitor cookie, and retries. If datacenter IPs keep getting refused, it moves to residential IPs for the rest of the run, at no extra cost to you. Failures are reported per input rather than crashing the run.
• This reads undocumented web JSON, which Weibo can change without notice. If a response no longer has the expected shape, the input is reported as parse_error rather than writing wrong data.

Privacy and responsible use

This Actor reads only what Weibo shows any logged-out visitor. It does not log in, and takes no cookies or credentials.

To keep personal data to what an analysis needs:

• Post and comment authors come with their public account fields only: name, link, avatar, verification and follower count. No gender, bio or location for people who are not the subject of your request.
• Email addresses and phone numbers are removed from all text (posts, comments, bios) and replaced with [email removed] / [phone removed]. This is the input Mask emails and phone numbers (maskContactInfo), on by default. Turn it off only if you have a lawful basis to process contact details; the text is then kept exactly as posted.
• Exact GPS coordinates that Weibo attaches to some posts are dropped; only the province label ("发布于 河北") is kept.
• Profiles carry account information only: no gender, birthday, school or employer.

Public posts and comments are still personal data under laws such as China's PIPL and the EU's GDPR. You are responsible for having a lawful basis for collecting and using this data, for how long you keep it, and for respecting Weibo's terms. This is not legal advice.

FAQ

Why did an input come back blocked?

Weibo refused every retry, on fresh IPs and visitor cookies. It happens in waves; run it again later. If many inputs in a row are refused, the run stops early (see Stop after this many refused inputs in a row) and the rest come back not_processed, free.

Why is a post not_found?

It was deleted, made private, or restricted by Weibo; or the link is mistyped. Weibo's own message is in error.

Can I get comments for posts found by keyword?

Run the keyword search first, then pass the url of the posts you care about into Post URLs in a second run. This keeps keyword runs cheap.

Why is memory limited to 1 GB?

The Actor reads JSON over plain HTTP and needs little memory; 512 MB is the default. More would not make it faster, only more expensive.

Something broke or a number looks wrong

Open an issue on the Issues tab with the input and the run link.