Answers
Will scraping Xiaohongshu get my account banned?
Short answer
It can: accounts used by scrapers get rate limited, flagged for automated activity and sometimes banned for good, and the risk grows with volume. The safe way is to keep your accounts out of it; okdata reads public notes and comments without any Xiaohongshu login, at $0.02 per search page and $0.02 per page of comments.
Yes, and it is the most common way these projects end. A brand manager logs a scraper in with the company account to pull comments for a launch report. A few hundred requests later the app shows a warning about automated activity, search stops working, and in the worst cases the account is banned permanently. People report it in open-source scraper issue trackers every month, sometimes after nothing more than viewing notes and saving videos at speed. Nobody outside Xiaohongshu knows exactly where the line is, and it moves.
Keep accounts out of data collection
The rule that works: never collect data with an account you care about, and preferably with no account at all. okdata fetches public Xiaohongshu data on its side and hands you JSON. You send an API key, never a cookie, so your brand account and your personal account are not involved in any request.
- Note search: about 20 notes per page for any keyword.
- Note comments: about 10 comments per page, with likes and reply counts.
- Note detail: full text and images of one note.
A launch report, priced
| Step | Calls | Cost |
|---|---|---|
| Find 100 notes about the product | 5 search pages | $0.10 |
| 50 comments on each of the top 20 | 100 comment pages | $2.00 |
| Total | 105 | {{cost:xiaohongshu.notes.search5+xiaohongshu.notes.comments.list100}} |
No account warm-up, no proxies, no captcha service. Each page is one charge; stop paging and the charges stop.
Example
curl -G "https://api.okdata.online/v1/xiaohongshu/notes/comments/list" \
-H "Authorization: Bearer $OKDATA_API_KEY" \
--data-urlencode 'id=6650a1b2000000001e00abcd'
The note ID is the 24-character string after /explore/ in a note link.
Fair limits
- Only public data. You cannot read anything a logged-out visitor could not see.
- Replies to comments are not returned; each top-level comment carries its reply count.
- Xiaohongshu hides view counts, so
metrics.viewsis always null.
If you still want to scrape yourself
Open-source tools like MediaCrawler are fine for small research jobs. Use a spare account that you can afford to lose, keep the pace slow, run during normal hours, and stop for the day at the first captcha or warning. If the data feeds a weekly report or a product, per-call pricing is usually cheaper than replacing accounts. The word-of-mouth page shows how a comment report is put together.
Updated