🇨🇳 Zhihu Q&A Tracker — Hot List & Knowledge-Mining for China Research

Pull Zhihu (知乎) — China's Quora — as structured data: the daily hot list (热榜) plus keyword-driven Q&A search, each record carrying the question, top-answer excerpt, voteup count, view count and category.

Zhihu is where mainland China goes for long-form, expert-leaning answers — the highest-signal Chinese surface for why people think something, not just that a topic is trending. This actor turns Zhihu's hot list and content search into clean JSON for China social-listening, consumer research and brand-monitoring work, without you needing a Zhihu account, a Chinese phone number or a CN VPN.

Built for market researchers, consumer-insight teams, brand monitors and offshore China-watchers who want the Zhihu signal as rows in a dataset, not a screenshot.

---

What you can do with it

• Consumer & market research — mine high-voteup answers for genuine expert opinion on a product, category or brand in China.
• Brand monitoring — run keyword search on your brand / competitors and read the top-answer excerpt and voteup count to gauge sentiment and depth of discussion.
• Trend tracking / social listening — capture the daily hot list (mode: "hot") on a schedule to see what questions China is asking right now.
• Offshore China-watching — read long-form Chinese knowledge content surfaced and categorised, filterable by voteup quality threshold.

---

What you get per record

Field	Type	Description
question_id	string	Zhihu question ID
question_title	string	The question (verbatim Chinese)
question_url	string	Canonical zhihu.com/question/{id} URL
category	string	Canonical category classifier: tech / finance / business / education / career / lifestyle / health / entertainment / science / politics
answer_count	int | null	Number of answers on the question, when surfaced
view_count	int | null	Question view / heat metric, when surfaced
top_answer_excerpt	string | null	Up to 500 chars of the highest-voted answer (HTML stripped)
top_answer_author	string | null	Handle of the top answer's author (匿名用户 if anonymous)
top_answer_voteup_count	int | null	Voteup count of the top answer — Zhihu's strongest quality signal
is_hot	bool	True if the record appeared on Zhihu's hot list
created_at	string | null	Question creation time, when surfaced
data_source	string	Provenance — exact probe path used (e.g. api.zhihu.com/topstory/hot-list, zhihu.com/hot (initialData), zhihu.com/search)

Most metric fields are nullable on purpose — Zhihu's anti-bot wall means not every path exposes every field on every run.

---

Input

Parameter	Type	Default	Description
mode	string	both	hot = daily hot list only; search = keyword search only; both = hot list, then top up with keyword search
keywords	array	["人工智能","新能源汽车","投资"]	Search terms (Chinese gives native-quality hits). Ignored when mode: "hot"
categories	array	["tech","finance","business"]	Restrict output to these category slugs. Empty = no filter
limit	integer	30	Max Q&A records (1–500)
min_voteup	integer	0	Drop records whose top answer has fewer than N voteups (0 = keep all)
include_hot_only	boolean	false	When true, keep only records that appeared on the hot list
proxyConfiguration	object	RESIDENTIAL	Apify proxy. RESIDENTIAL strongly recommended — Zhihu blocks datacenter IPs

Sample input

{
  "mode": "both",
  "keywords": ["人工智能", "新能源汽车", "投资"],
  "categories": ["tech", "finance", "business"],
  "limit": 30,
  "min_voteup": 0,
  "include_hot_only": false,
  "proxyConfiguration": {
    "useApifyProxy": true,
    "apifyProxyGroups": ["RESIDENTIAL"]
  }
}

Sample output (truncated, schema-accurate)

[
  {
    "question_id": "650000001",
    "question_title": "如何看待新能源汽车 2026 年的价格战？",
    "question_url": "https://www.zhihu.com/question/650000001",
    "category": "finance",
    "answer_count": 412,
    "view_count": 3800000,
    "top_answer_excerpt": "价格战的本质是产能过剩下的份额博弈……（节选）",
    "top_answer_author": "汽车行业分析师",
    "top_answer_voteup_count": 5821,
    "is_hot": true,
    "created_at": "2026-04-02T09:15:00Z",
    "data_source": "api.zhihu.com/topstory/hot-list"
  }
]

---

How it gets the data

Zhihu is mainland-hosted and applies aggressive anti-bot detection. Route status verified live on 2026-09-05 from a datacenter IP:

1. Mobile API hot list — api.zhihu.com/topstory/hot-list — HTTP 200, works without a proxy. This is the primary route for mode: "hot".
2. Public hot-list pages — zhihu.com/hot, zhihu.com/billboard — HTTP 403 (anti-bot interstitial) from datacenter IPs; retried behind residential CN sessions.
3. Public content search — zhihu.com/search?type=content — HTTP 403 from datacenter IPs; this is the route for mode: "search" and it needs a residential (ideally CN) exit.
4. Mobile API search — api.zhihu.com/search_v3 — returns need_login (HTTP 403) and is no longer used.

Every blocked request is retried up to 4 times, each attempt on a fresh residential proxy session (CN → HK → SG). If every attempt is blocked, the run FAILS with a status message naming the cause. It does not emit a placeholder row, and you are not charged.

FAQ

Can I scrape Zhihu without an account or Chinese phone number? Yes. The actor runs server-side behind Apify proxies and returns JSON — no Zhihu login, phone number or VPN on your side.

What is Zhihu? Zhihu (知乎) is China's largest long-form Q&A / knowledge community — the closest mainland equivalent to Quora, but with far deeper expert participation.

How do I get only the trending hot list? Set mode: "hot" (or include_hot_only: true) to keep only hot-list questions and drop keyword-search noise.

How do I filter for high-quality answers? Use min_voteup — 100+ usually means substantive expert content, 1000+ usually means a viral / canonical answer.

How fresh is the data? Each run captures the hot list / search live and the record's data_source shows the exact path used. The actor has no internal scheduler — schedule it in Apify for a continuous feed.

---

Related actors — NexGenData China social-listening fleet

Pair Zhihu with the rest of the NexGenData Chinese-social fleet for full cross-platform coverage:

• Weibo Hot Search Tracker — China's #1 social-trending barometer (微博热搜榜): the fast real-time counterpart to Zhihu's long-form signal.
• RedNote (Xiaohongshu) Scraper — trending posts, feeds and notes from Xiaohongshu (小红书 / RedNote) for beauty/fashion/lifestyle consumer intent.
• Bilibili Video Search — keyword search across Bilibili (B站), the long-form video, gaming and education hub of mainland China.
• Douyin Trending Tracker — the live Douyin (抖音 / Chinese TikTok) hot list for short-video trend signal.
• Kuaishou Trending Tracker — trending short-video signal from Kuaishou (快手), skewing lower-tier-city audiences.
• Douban Tracker — China movie/TV ratings and hot lists from Douban (豆瓣), the cultural-taste signal.
• China Trends Tracker — cross-platform Chinese trend roll-up (Weibo / Baidu / Toutiao / Douyin).
• Chinese Social Signals MCP — MCP server that plugs the whole Chinese-social fleet directly into Claude, ChatGPT and Cursor.

---

Notes & limits

• Residential proxy strongly recommended. Datacenter IPs are blocked / captcha-gated; the default proxy group is RESIDENTIAL.
• No placeholder rows, ever. If Zhihu blocks every path the run ends FAILED with an explanatory status message and zero dataset rows — you are never charged for a diagnostic row. If Zhihu answers but nothing matches your filters, the run SUCCEEDS with 0 rows and a message saying so.
• Excerpt is capped at 500 characters of the highest-voted answer, HTML stripped.
• Pay-per-event billing. $0.10 per delivered Q&A record, plus $0.01 per actor start. Rows are charged only when a real record is written to the dataset — errors, blocks and empty results cost nothing beyond the start event.
• Run budget. The actor derives a deadline from the run timeout and stops cleanly before it, reporting "stopped early at the run time limit". Partial runs are never silent.
• Top-answer fields. top_answer_excerpt comes from the hot list's answer preview on mode: "hot"; top_answer_author / top_answer_voteup_count are only surfaced by the search route and are null on hot-list records rather than filled with a guess.