Can an AI agent shortlist Xiaohongshu or Douyin creators for a campaign?

A skincare brand needs 20 mid-tier Xiaohongshu creators, 10,000 to 100,000 followers, who post about sensitive skin (敏感肌), whose followers actually engage, and whose image note costs no more than ¥4,000. An agent can do the legwork in an afternoon: find who is booked for this topic, check their numbers and prices, read their latest notes, and hand you a ranked table with links and a sentence on why each one fits. You still make the call; the agent saves you the 200 profile visits.

How the agent finds creators

1. Search the marketplace. Xiaohongshu PGY creator search with the topic as query, followers_min=10000, followers_max=100000 and sort=most_engagements. Each page has 20 creators with followers, median reads and engagements of everyday notes, content categories and what an image note and a video note cost in yuan.
2. Drop the misfits. Over budget, obvious brand accounts, categories that do not match.
3. Read recent work. The PGY creator ID is the Xiaohongshu user ID, so Xiaohongshu account notes takes it as is. The latest titles show whether sensitive skin is a habit or a one-off.
4. Check finalists. Xiaohongshu PGY creator detail adds likes and saves, content and feature tags, and whether each note type is open for orders now.
5. Judge fit. The LLM compares recent titles and tags with your brief, works out the price per 1,000 median reads, writes one line per creator and sorts.

Good creators who never joined PGY will not show up. To catch them, the agent can also run Xiaohongshu note search and look at who writes the top notes.

For Douyin, swap in Douyin Xingtu creator search: same follower bounds plus engagement_min, with 30-day engagement rate, median views and the integrated and custom video prices in every result. The Xingtu rate card gives finalists' full price list.

What a shortlist costs

Step	Calls	Cost
3 PGY searches, 60 creators with rates	3	$0.12
Recent notes for 40 creators	40	$0.80
PGY detail for 20 finalists	20	$0.80
Total, Xiaohongshu	63	{{cost:xiaohongshu.pgy.creators.search3+xiaohongshu.users.notes.list40+xiaohongshu.pgy.creators.get*20}}
Douyin: 3 Xingtu searches and 20 rate cards	23	{{cost:douyin.xingtu.creators.search3+douyin.xingtu.creators.rates.get20}}

Code

import os, requests

API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}

def get(path, **params):
    body = requests.get(API + path, headers=H, params=params, timeout=60).json()
    if body["code"] != "ok":
        raise RuntimeError(f"{body['code']}: {body['message']}")
    return body["data"]

def image_note_price(c):
    return next((r["price"] for r in c["rates"] or [] if r["type"] == "image_note" and r["price"]), None)

found = {}
for kw in ["敏感肌 修护", "敏感肌 面霜", "换季 泛红"]:
    for c in get("/v1/xiaohongshu/pgy/creators/search", query=kw, followers_min=10000,
                 followers_max=100000, sort="most_engagements")["items"]:
        found[c["id"]] = c

shortlist = []
for c in found.values():
    price = image_note_price(c)
    if not price or price > 4000:
        continue
    notes = get("/v1/xiaohongshu/users/notes/list", id=c["id"], fields="title")["items"]
    shortlist.append({"name": c["name"], "url": c["url"], "followers": c["followers"],
                      "image_note_cny": price, "median_reads": c["median_views"],
                      "categories": c["categories"], "recent": [n["title"] for n in notes[:10]]})
# hand `shortlist` and your brief to the LLM: "rank by fit, one line each on why"

Over MCP, an agent can run the same steps with call; ask it to state how many creators it will check before it starts.

What the data still cannot tell you

• Audience demographics. Follower age, gender and city split stay in the PGY and Xingtu consoles, which is also where you book.
• Final prices. Rates are what creators list; fees and discounts are settled when you order on the platform.
• Bought engagement. High likes with almost no comments is a warning sign. Pull a page of note comments on two or three notes before you commit.

Other ways

Agencies and influencer platforms come with vetted lists and contracts, which matters for a large campaign. Tools such as 千瓜 or 蝉妈妈 rank creators by category without code. An agent working from marketplace data is the quick first pass that tells you who to look at and what they will cost. If all you need is a filter by follower band and budget, see shortlisting creators by follower range and price. More on this job: creator discovery.