DEV Community

Cover image for Influencer Vetting at Scale on Xiaohongshu (RedNote): A Practical Python Guide for Brand Teams 2026
Sami
Sami

Posted on

Influencer Vetting at Scale on Xiaohongshu (RedNote): A Practical Python Guide for Brand Teams 2026

#python #webscraping #china #marketing

RedNote — known internationally as Xiaohongshu (小红书) or Little Red Book — has become the single most consequential platform for influencer marketing in China. With 300M+ monthly active users skewed female and Gen Z, it's where beauty, fashion, lifestyle, and travel brands first place a campaign before going wider. After the TikTok-uncertainty migrations of 2024–2025, RedNote also became the de facto Western fallback for many creators.

If you're a brand team, agency, or media buyer working in or with China, you need a way to vet RedNote influencers at scale. Manual scrolling doesn't cut it past five creators. Here's the structured approach.

What "vetting at scale" actually means

For a single influencer partnership, you typically want answers to:

  • Reach: how many followers, but more importantly how many people actually see their content (median impressions / followers)?
  • Engagement quality: average likes/comments/saves per post — and the distribution. A creator with 100K followers and 50 posts averaging 500 likes is very different from one with 100K and a few viral 10K posts.
  • Niche fit: do their tags and topics align with your category?
  • Audience hints: from bio, location, profile signals — who follows them?
  • Authenticity: posting cadence, sponsored-content ratio, content reuse from other platforms.

The data lives in two places on RedNote: profile metadata (followers, bio, location, verified status, total likes received) and recent posts (titles, like counts, content type, publish dates).

What you can extract per creator 

{
    "userId": "5d7439b40000000001009f54",
    "nickname": "BeautyBlogger123",
    "avatar": "https://sns-avatar-qc.xhscdn.com/...",
    "redId": "100123456",
    "description": "Skincare reviews, K-beauty translations. Seoul-based.",
    "followers": 184500,
    "following": 320,
    "totalLikes": 1240000,
    "gender": 1,
    "location": "Seoul, South Korea",
    "isVerified": true,
    "tags": ["Beauty Blogger", "K-Beauty", "Skincare"]
}
Enter fullscreen mode Exit fullscreen mode

Per recent post:

{
    "postId": "64be395b0000000010030b56",
    "type": "video",
    "title": "Morning skincare routine for dry skin",
    "likes": 15234,
    "scrapedAt": "2026-05-06T12:00:00Z"
}
Enter fullscreen mode Exit fullscreen mode

Combined, you can compute:

  • Median likes / post (use median, not mean — viral outliers skew means)
  • Engagement rate = median likes / followers (RedNote benchmark for healthy: 2–5%, viral creators 8%+)
  • Post frequency (posts per week — burnout warning if dropping)
  • Content-type ratio (video vs image — videos get higher reach in 2026)
  • Niche concentration (% of recent posts matching your category keywords)

Practical Python: building a vetting batch 

from apify_client import ApifyClient
import statistics

client = ApifyClient("YOUR_APIFY_TOKEN")

def vet_creator(profile_url: str) -> dict:
    # Profile data
    profile_run = client.actor("zhorex/rednote-xiaohongshu-scraper").call(run_input={
        "mode": "profile",
        "userUrl": profile_url,
    })
    profile = next(client.dataset(profile_run["defaultDatasetId"]).iterate_items())

    # Recent posts
    posts_run = client.actor("zhorex/rednote-xiaohongshu-scraper").call(run_input={
        "mode": "user_posts",
        "userUrl": profile_url,
        "maxResults": 50,
    })
    posts = list(client.dataset(posts_run["defaultDatasetId"]).iterate_items())

    likes = [p.get("likes", 0) for p in posts if p.get("likes")]
    median_likes = statistics.median(likes) if likes else 0
    er = (median_likes / profile["followers"]) * 100 if profile.get("followers") else 0

    return {
        "nickname": profile.get("nickname"),
        "followers": profile.get("followers"),
        "verified": profile.get("isVerified"),
        "tags": profile.get("tags"),
        "post_count_recent": len(posts),
        "median_likes": median_likes,
        "engagement_rate_pct": round(er, 2),
        "video_ratio": sum(1 for p in posts if p.get("type") == "video") / max(len(posts), 1),
    }

# Vet a list
candidates = [
    "https://www.xiaohongshu.com/user/profile/USER_ID_1",
    "https://www.xiaohongshu.com/user/profile/USER_ID_2",
]

results = [vet_creator(u) for u in candidates]
results.sort(key=lambda r: r["engagement_rate_pct"], reverse=True)

for r in results:
    print(f"{r['nickname']:25s}  followers={r['followers']:>7,}  ER={r['engagement_rate_pct']}%")
Enter fullscreen mode Exit fullscreen mode

That gives you a sortable spreadsheet of candidates ranked by genuine engagement, not vanity follower counts.

Red flags to look for

  • Engagement rate < 0.5% with > 100K followers → likely bought or stale
  • Only image posts when category is video-heavy → low reach in 2026's algorithm
  • Posts in the last 30 days < 4 → low retainer reliability
  • No tags or generic tags → low discoverability inside RedNote search
  • Median likes within 10x of mean → relatively consistent (good); 50x+ means single-viral-driven (risky)

When to skip the DIY approach

RedNote's public surface requires sustained anti-bot engineering to extract reliably at scale. The schema also evolves regularly, so a scraper that worked last month can quietly start returning empty arrays this month without raising any error you'd notice in your pipeline.

That's why I maintain the RedNote Scraper on Apify — six modes (search, user posts, profiles, post details, comments, video) with consistent output schemas across them. The infrastructure work (session handling, rate limiting, schema stability) is already done.

Pricing is pay-per-event: $0.005 per result. A typical influencer batch (50 candidates, profile + 50 posts each) costs about $13. The Apify free tier ($5 monthly) covers ~1,000 items.

FAQ

Is the data from open profiles only? Yes — public-facing profile and post data. Private/locked accounts are not accessible. Same content any anonymous browser user can see.

Does it work on xhslink.com short links? Yes, those resolve automatically.

Can I get post-level comments? Use mode: comments on individual post URLs.

What about RedNote vs Xiaohongshu vs Little Red Book? All the same platform — the Actor handles all three name conventions and URL formats.

Is scraping Xiaohongshu legal? This Actor accesses publicly visible content only. No authentication is bypassed. Always consult your local laws.

Full Chinese intelligence stack

Brand teams running campaigns across Chinese platforms typically pair this with:

Vetting more than 100 creators per month? I offer custom output schemas, dedicated proxy pools, SLA support, and volume discounts. DM me on Apify or open an Issue titled "Enterprise inquiry".

Bug? Issues are typically fixed within 48 hours.

If this saved you time, a 30-second review on the Apify Store helps a lot. ⭐

Top comments (0)