Zhihu (知乎) and Bilibili are both described abroad as consumer platforms, which is why they get skipped. Zhihu is read as a Chinese Quora, Bilibili as a Chinese YouTube for cartoons. Neither description survives half an hour spent on either site looking for an answer to a technical question. What you find instead is the place where a specification gets discussed by people who use it, where a category gets compared brand by brand, and where the sentence a buyer will repeat to their own management was first written by someone else.
Two platforms, two reading contracts
On Zhihu, the unit is the answer, and the question it answers has usually been asked by a stranger with a real problem. An answer that is upvoted rises inside that question and stays there; the question itself keeps collecting readers long after the discussion has gone quiet. Authority is granted by demonstration rather than by title: a long answer that explains why a material behaves the way it does, with the reasoning visible, outranks a short answer from someone whose profile says they are an expert. The platform also carries longer columns, which behave more like articles and less like answers, and which are the right place for something that nobody has asked yet.
Bilibili is built on long video, and that single fact changes everything about what is possible there. A viewer who has decided to watch will stay for a demonstration that takes its time: a teardown, a test rig, a comparison filmed from start to finish, a shop-floor walkthrough. The audience skews young and technical, and it is unusually willing to watch someone be precise. Comments scroll across the image as the video plays, which means the audience annotates the demonstration in real time — including the moment where you cut a corner. Nothing is more exposed, and nothing builds more credit when the demonstration holds.
| Zhihu | Bilibili | |
|---|---|---|
| What a piece of content is | An answer inside a question someone else asked, or a column article | A video long enough to show a whole operation, with its description and pinned comment |
| Where the authority comes from | The reasoning being visible and checkable by readers who do the same work | The demonstration being filmed without cuts at the moment that matters |
| How a reader finds you later | Through the question, which keeps being searched and read long afterwards | Through the platform's own search and recommendation, and through saved and collected videos |
| What kills a piece immediately | A conclusion stated without the reasoning that would allow disagreement | A promise in the first seconds that the rest of the video does not keep |
Why a promotional article dies there
On both platforms the audience votes, and on both platforms the vote is read by the ranking. A text that opens on what the company is proud of, lists advantages, and ends on an invitation to get in touch has no mechanism by which it can rise: nobody upvotes an argument they were not allowed to examine. Worse, the audience names the genre out loud. A piece recognised as advertising gets said so in the comments, in the first few, and the label travels with the content from then on. The brochure is not merely ignored, it is identified — and the identification is visible to the next reader.
The way out is not to hide the commercial interest, which fools nobody and reads worse when it is found. It is to accept the platform's bargain: you give away the reasoning, and in exchange the reader decides for themselves. Say what your approach does badly. Name the case where a competing method is the right choice. Explain the trade-off you made and why someone with different constraints would make the other one. A reader who has been given the means to disagree and ends up agreeing has become something a brochure cannot produce.
What is worth publishing, and in which tone
- The answer to a question that is already being asked. Search the platform first for how your category is discussed, in the words buyers use rather than the words your catalogue uses, and answer the question that already has readers instead of opening a new one nobody will find.
- A method, written so that it can be reproduced. How you test something, what the test does not cover, what you would measure differently next time. A method is useful even to someone who will never buy from you, which is exactly why it circulates.
- A failure, described without softening it. The part that cracked, the assembly that did not hold, what the analysis found. This is the single most credible thing a company can publish, and almost nobody publishes it.
- A comparison you did not rig. If your product loses on one criterion, say which and by how much in words. A comparison in which you win everywhere is read as a comparison you wrote, because it is.
- What the European practice is, and why. A Chinese technical reader is genuinely curious about how a problem is handled elsewhere. This is the one subject on which a foreign company has something nobody local can write.
The tone that works on both platforms is the one you would use explaining the subject to a competent colleague who does not know this particular file. Flat, specific, unhurried, with the uncertainties left in. It is further from marketing writing than most companies expect, and closer to the internal note that was never meant to be published — which is a useful instinct when deciding what to write.
An engineer with a name beats a brand account
A company account on either platform is reading itself out loud. Everything it publishes is understood to have been approved, which means it can never say the one thing that would make it credible: this is what I think, and here is where I am not sure. An individual account held by someone who actually does the work can say both. It can answer a comment in its own voice, admit it was wrong about something, and take a side in an argument — none of which a brand account is structurally able to do.
That the person is identified as working for the company is not a problem, provided it is stated rather than discovered. Readers are entirely comfortable with an engineer who says where they work and then explains something properly; they are not comfortable with a neutral-looking account whose affiliation turns up in the comments. The arrangement that holds is simple: the person owns the account and the voice, the company owns nothing but the fact of being mentioned, and nobody reviews a comment before it is posted. A company that cannot live with that should publish on its own site instead, where the rules are its own.
How long a good piece keeps working
This is the argument that usually decides whether the effort is worth making. A post on a feed-driven network is read on the day it is published and then effectively gone. A good answer on Zhihu sits inside a question that people keep asking, and it keeps being read for as long as the question is live — which, for a technical question, is a long time. A demonstration video on Bilibili is found by search and by recommendation, and the ones that explain something properly get collected and sent to colleagues, which revives them periodically. The practical consequence is that the right rhythm is not frequent but durable: fewer pieces, each one answering something precise, each one written so that it is still correct a year later. Publishing six things a month on these two platforms is a way of wasting a year; publishing something properly finished every few weeks, and answering every serious comment underneath it, is how a foreign name becomes one a Chinese engineer has read before.
Before writing anything, spend an afternoon reading. Search your own category on both platforms and note the questions that keep coming back, the objections that get repeated, and the words buyers use for what you sell. That list is worth more than a content plan, because it is the only version of your market written by the people in it.