Substack and YouTube educators know the temptation. Start with a public question, give a useful answer, add a card for the paid product and assume the content has become a funnel.
Zhihu’s public product surfaces make the sequence visible, but they do not make the conclusion true. The official creator interface has a question-search entry. A Zhihu FAQ describes creating a paid column from the creator centre and inserting a column card in an answer. These are different product objects and flows. They do not prove that an answer will be discovered, that a reader wants the paid column, or that a creator is eligible to use every feature.
The useful decision for an overseas expert is not “How do I make Zhihu convert?” It is “Which content object should answer which reader question, and what evidence do I still need?”
Public answer and paid column solve different jobs
A public answer can clarify a question, show how the expert thinks, and reveal which terms or assumptions need explanation. It may be useful before a product description is final.
A paid column packages a set of content as a product object. It may have a title, description, sequence and card that can be inserted into an answer. The FAQ describes that product flow. It does not say the card will be shown to the right reader, that the reader will purchase, or that the column fits an overseas course.
Treat the answer as an explanation surface and the column as a product surface. A possible link between them is a design choice, not a verified funnel.
Read the two official pages narrowly
Zhihu’s question-search/creator-centre page establishes that question search is an available interface entry in the page observed on 2026-08-23. It does not establish that a particular question has demand, that an answer will receive exposure, or that an overseas organisation can register or publish.
The paid-column FAQ describes a creation flow and the possibility of inserting a paid-column card in an answer. It does not establish creator qualification, pricing, revenue, payment success, refund behaviour or user willingness to pay.
These limits are not a reason to ignore the pages. They are the reason to use them as interface evidence rather than as a business forecast.
A question-led workflow for an expert product
First: name the reader’s question
Start with a question the intended reader is genuinely trying to clarify. “How do I build an AI strategy?” is broad. “How can an operations lead decide whether an AI workflow needs a retrieval check?” gives the answer a visible decision.
The question search page can help an editor inspect language and context. It cannot prove that the question represents a buyer segment. Record it as a research input.
Second: write a useful public answer
The answer should stand on its own. It should explain one problem, state what the source or method can and cannot establish, and give the reader a next question. Do not make the answer a teaser that withholds the only useful point until payment.
The public answer is also a test of language. If readers misunderstand the work being described, the expert may need to revise the product promise before packaging a paid column.
Third: decide whether a paid column is the right object
A paid column is not automatically the next step. If the expert has a defined sequence, audience, permission to use the material and a clear promise, a paid product may be worth preparing. If the answer is still testing the problem statement, keep the work in research or public explanation.
The FAQ supports the existence of a card-in-answer flow. It does not tell the expert when the product is ready.
Fourth: preserve the gap between card and conversion
If a card is inserted, describe it as an available navigation or product-presentation action. Do not call it a conversion, sale, qualified lead or demand signal without separate evidence. A card can be visible and still not be clicked. A click can happen without payment. Payment can happen without a completed learning result.
A scene: the answer that tried to sell too early
An overseas expert writes a public answer about AI workflow evaluation. The answer contains a paid-column card titled “Build Reliable AI Agents.” The expert reports that the funnel is live.
What is actually known? A public answer was written. A card was inserted according to the documented product flow. It is not yet known whether the question matched the reader, whether the card was seen, whether anyone paid, or whether the column delivered the intended learning result.
A safer editorial record says: “The answer tests the problem language. The column card presents a possible next product. Results and qualification are not established.” This is a constructed example, not a reported platform result.
The product needs a boundary before the card
Before building the paid column, list the promise, intended reader, source material, sequence and review condition. If the column reuses an expert course or paid material, authorization and rights remain separate from the platform flow. A card cannot grant permission to translate, publish or distribute someone else’s content.
Also state what the product does not promise. It may explain a method without guaranteeing a job result. It may provide a sequence without promising certification. It may help a reader frame a problem without proving a China market opportunity.
A counterexample: when the paid column can come first
If an expert already has a defined product, an approved content scope, a known audience and a clear delivery plan, creating the paid column before broad public question research may be reasonable. This is a product-preparation decision. It still does not prove discovery or payment.
Likewise, a public answer may not be needed for a private, invited audience. Do not force a public funnel onto a product whose reader and delivery are already defined.
A three-state content record
For each question/product pair, record:
- Question evidence: what wording and context were observed.
- Answer evidence: what the public explanation actually clarified.
- Product evidence: what the paid column contains and what remains unverified.
Do not add “conversion” as a fourth state until an actual, permissioned measurement exists. Even then, conversion is not learning impact or business success.
The sentence worth carrying forward is: a paid-column card can connect two content objects; it cannot certify a funnel.
If you need to decide whether an expert product is ready to move from a public question to a paid content object, request a China validation review. Bring the question, answer and product promise separately.
Related reading
- On Zhihu, Start With the Question—not the Content Calendar
- A Product Portability Audit Comes Before China Distribution
- WeChat After Interest: Service Continuity for Expert Products
- When Topic Popularity Is Not Buyer Demand
Sources and evidence boundary
- Zhihu question search / creator centre, checked 2026-08-23. Shows an official question-search entry; does not establish demand, exposure or eligibility.
- Zhihu paid-column FAQ, checked 2026-08-23. Describes paid-column creation and answer-card insertion flow; does not establish qualification, revenue, payment or conversion.
Image brief: Two editorial cards: a public question-led answer and a paid-column product card, connected by a possible review step rather than a conversion arrow. No revenue figures, account badges or platform-growth claims. Alt text: “A public Zhihu answer and a paid column shown as separate content objects.”