How to scrape Douyin videos, comments, and creator profiles

Which Douyin URLs can you scrape, and what does each one return?

Datapika reads three Douyin URL shapes and detects the platform from the domain, so you can leave the platform setting on auto for douyin.com and iesdouyin.com links. A video URL in the form douyin.com/video/<id> feeds video_detail and video_comments. A creator URL in the form douyin.com/user/<sec_uid> feeds user_posts, user_profile, and user_likes, and a live.douyin.com/<room_id> URL feeds live_info.

The trending mode needs no URL, but it does need platform set to douyin explicitly. Left on auto, trending falls back to Bilibili's popular feed instead. Douyin's trending surface is a board of ranked search phrases with a hot_value score, not a video feed, so those rows carry item_type hot_search_keyword rather than video.

Some newer short-form Douyin share links are not parsed yet. If a share link fails, open it once in a browser and paste the canonical douyin.com/video/<id> form instead. Every unsupported input still produces a structured error row instead of a failed run.

• douyin.com/video/<id>: single video detail, or its comment thread with replies
• douyin.com/user/<sec_uid>: latest posts, profile card, or liked videos for one creator
• live.douyin.com/<room_id>: live status, stream title, viewer count, host name, cover image
• No URL plus platform=douyin: the hot-search keyword board with hot_value scores
• iesdouyin.com links are recognized as Douyin automatically
• Unparsed short links return an error row; paste the canonical video URL to retry

Which fields does a Douyin video, comment, or profile row include?

A Douyin video row carries the first line of the caption as title (capped at 200 characters) and the full caption as description, the aweme ID as item_id, a canonical douyin.com/video URL, the creator's sec_uid as author_id, nickname as author_name, and Douyin ID as author_username. It adds duration_sec, cover_url, music_title, music_author, created_at as an ISO 8601 timestamp, and a hashtags array whose extractor handles Chinese characters. Engagement comes back as play_count, like_count, comment_count, share_count, and collect_count, the last one being Douyin's save-to-favorites metric.

A creator profile row adds follower_count, following_count, video_count, bio, avatar_url, and a verified flag that is true for both personal and enterprise verification badges. Comment rows include comment_id, reply_to_id for threading, comment_text, the commenter's identifiers, like_count, and created_at.

Every row also includes the raw source response under raw and a scraped_at timestamp, so Douyin fields that are not mapped into the unified schema are still available for analysis.

• Video: title, description, hashtags, music_title, music_author, duration_sec, cover_url, created_at
• Engagement: play_count, like_count, comment_count, share_count, collect_count
• Profile: follower_count, following_count, video_count, bio, avatar_url, verified
• Comment: comment_id, reply_to_id, comment_text, like_count, created_at
• Live: live_status, live_title, viewer_count, author_name, cover_url
• raw and scraped_at on every row for unmapped platform fields and audit trails

How much does scraping Douyin cost, and how do bulk runs work?

Datapika charges $0.001 per delivered record as of August 2026 (Apify Store), which is $1.00 per 1,000 rows. Douyin URLs that fail because a video was deleted, a creator went private, or the platform blocked a request return an error row with a readable error_message and are not charged. There is no subscription and no API key fee on top of the per-record price.

For bulk work, pass a list under urls instead of a single url. URLs are processed concurrently, and the maxItems setting (1 to 5,000, default 100) caps how many rows each paginated URL returns, so 20 creator URLs at maxItems 50 yields up to 1,000 video rows for about $1.00. A batch of 100 URLs typically finishes in 2 to 3 minutes, and runs of 500 URLs or fewer are the most reliable.

You can also cap spend for a run with Apify's maximum total charge setting. When the cap is hit mid-batch, remaining URLs are written as error rows and the run ends cleanly instead of overspending.

• $0.001 per delivered record (Apify Store, August 2026); the actor has 1,340 runs
• Error rows are free: deleted, private, region-blocked, or anti-bot failures cost nothing
• maxItems from 1 to 5,000 per input URL, default 100
• About 100 URLs per run in 2 to 3 minutes; keep runs under 500 URLs for best reliability
• Spending cap: remaining URLs become error rows once the maximum total charge is reached

What can you build with Douyin data for the Chinese market?

Douyin is the domestic Chinese counterpart to TikTok, and its data answers questions that Western platforms cannot. Brands selling into China use user_posts across competitor accounts to watch launch cadence, caption language, and collect_count as a save-intent signal. Agencies running KOL discovery pull user_profile rows for a candidate list and rank by follower_count against average play_count from the same creators' recent posts.

The hot-search board is a fast read on what mainland audiences are searching right now. Scheduling the trending mode every hour and diffing hot_value across snapshots surfaces rising phrases before they show up in English-language coverage. Comment threads in Chinese feed sentiment models directly, since comment_text and like_count come back per reply with threading intact.

Because the schema is shared with Datapika's TikTok and Bilibili modes, the same brand can be compared across all three platforms in one dataset without remapping fields. That matters for teams that report on China and global audiences side by side.

• Competitor launch tracking: user_posts on rival accounts, sorted by created_at and collect_count
• KOL discovery: user_profile rows ranked by follower_count and recent play_count
• Hot-search monitoring: hourly trending snapshots with hot_value deltas
• Chinese-language sentiment: threaded comment_text with like_count per reply
• Live commerce checks: live_info for viewer_count and stream titles during campaign windows
• Cross-platform reports: identical field names across Douyin, TikTok, and Bilibili

How do AI agents call the Douyin scraper through the API or MCP?

The actor is exposed on Apify as a standard REST endpoint, so any HTTP client can start a run with a JSON body containing mode, url or urls, and maxItems, then read the dataset as JSON or CSV when the run finishes. Authentication is your Apify token only; no Douyin credentials or cookies are required for public content.

For agents, the same actor is available as an MCP tool through Apify's MCP server at mcp.apify.com. An agent connects with the actor enabled as a tool, calls it with the Douyin URL and mode as arguments, and receives the unified rows back in the tool result. Billing stays per record, so an agent that asks for 30 comments pays $0.03.

The cookie field is optional and only needed for gated Douyin content such as private collections or certain feeds. Residential proxy is the default and the recommended choice for Douyin. When Douyin flags an exit IP, the actor retries on a fresh residential IP up to 4 times before writing an error row.

• REST: POST a JSON input to the actor endpoint, then fetch the dataset items
• MCP: enable openclawai/tiktok-douyin-bilibili-scraper as a tool on mcp.apify.com
• Inputs: mode, url or urls, maxItems, optional includeComments, cookie, proxyConfiguration
• Public content needs no cookie; gated collections accept a pasted browser cookie
• Flagged exit IPs are retried on a fresh residential IP up to 4 times per URL

Questions, answered

Keep reading