# Why can't Perplexity or ChatGPT search find Douyin and Xiaohongshu posts?

> Douyin and Xiaohongshu keep most posts out of web search indexes, so AI search tools miss them. Give your agent direct search on both platforms. Costs and code.

Canonical page: https://okdata.online/answers/ai-search-cant-find-douyin-xiaohongshu/

**Short answer: **Douyin and Xiaohongshu keep most of their content inside their apps, behind logins and JavaScript, so search engines index little of it and AI search built on those indexes rarely cites it. To include them, give the agent each platform's own search: a Douyin video search costs $0.02 and a Xiaohongshu note search $0.02, each for a page of about 20 results.

Ask Perplexity or ChatGPT search what Chinese shoppers say about a sunscreen this summer and the answer cites news sites, Zhihu threads and the brand's own pages. The thousands of reviews on Xiaohongshu and the try-on videos on Douyin barely show up, even though that is where the conversation happens. Both platforms keep their content in the app, put a login in front of the web version and render posts with scripts, so web crawlers see little and search indexes stay thin. AI search tools mostly read from those indexes, so the gap carries straight into their answers.

## Give the agent the platforms' own search

Keep web search for news and articles, and add two tools for the social side:

- [Douyin video search](/apis/douyin/videos/search/): keyword in, about 20 videos out, with likes, comments, shares and the author; sort by relevance, latest or most liked.
- [Xiaohongshu note search](/apis/xiaohongshu/notes/search/): about 20 notes per page with text, images, likes and saves; sort general, latest or popular, filter to video or image notes.

With the [MCP server](/mcp/) connected, an agent can do this itself. In code, run both and label the results so the model can tell sources apart:

```python
import os
import requests

API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
FIELDS = "url,title,text,published_at,metrics.likes,metrics.comments,author.name"

def search(path, query, sort):
    body = requests.get(API + path, headers=H, timeout=60,
                        params={"query": query, "sort": sort, "fields": FIELDS}).json()
    if body["code"] != "ok":
        return []
    return body["data"]["items"]

query = "防晒霜 测评"
results = (
    [{"platform": "douyin", **p} for p in search("/v1/douyin/videos/search", query, "most_liked")]
    + [{"platform": "xiaohongshu", **p} for p in search("/v1/xiaohongshu/notes/search", query, "popular")]
)
print(len(results), "posts to add next to the web results")
```

## What a research question costs

| Task | Calls | Cost |
|---|---|---|
| One keyword, one page on each platform | 2 | $0.04 |
| 3 keywords, 2 pages each, both platforms | 12 | $0.24 |
| Weekly scan of 10 keywords, both platforms | 20 | $0.40 |

## Gotchas

- Search in Chinese. Most posts are written in Chinese, and an English keyword finds a small slice.
- On Douyin the text is the caption, not what is said in the video. Captions are short, so read likes and comments as the signal and fetch comments for depth.
- Xiaohongshu hides view counts; that field is always null. Saves and likes tell you more.
- Ranking is the platform's own and changes over time. Store what you found with the date.
- Public posts only.

## When to do it differently

Both apps now have AI search built in. For a single question asked by hand, typing it into Xiaohongshu or Douyin on your phone is free and good. Chinese social analytics dashboards suit teams that want charts rather than raw posts. An agent with platform search earns its keep when the question repeats, spans several platforms or feeds a report. See [content research](/solutions/content-research/) for the wider set of endpoints.

## Endpoints used

- [Douyin video search](https://okdata.online/apis/douyin/videos/search/) `douyin.videos.search`: $0.02 per call · about 20 videos per page
- [Xiaohongshu note search](https://okdata.online/apis/xiaohongshu/notes/search/) `xiaohongshu.notes.search`: $0.02 per call · about 20 notes per page

Updated 2026-11-12