Douyin Search: The Discovery Layer Treated as an Afterthought
Douyin reads the words inside your video, not just the caption. Footage with no Chinese anywhere in it is invisible to a Chinese query, however good it looks.
Most overseas brands think about Douyin as something that happens to them. You post, the feed either picks it up or it doesn't, and the outcome feels like weather. That describes half the app. The other half is a search box, used by people who have already decided to travel and are now looking for specifics, and it behaves far more like a search engine than like a feed. It is also the half that overseas accounts tend to leave alone, which is odd, because it's where intent lives.
The short version: Douyin doesn't only match your caption and hashtags. It is commonly described as reading the text on screen and transcribing what is spoken, which would make the searchable surface of a video everything written and said in it. A beautifully shot clip with an English caption, no subtitles and a music bed contains no Chinese for a Chinese query to match. Fixing that costs nothing but attention: write the caption in the words a traveller would actually type, burn Chinese subtitles, and give your business a Chinese name people can type in the first place.
The half with intent in it
The feed-versus-search split works the same way here as it does on Xiaohongshu, and we've already laid it out in why good notes keep selling for months: one surface interrupts, the other answers a typed question, and only the second keeps being retrieved once the video has stopped being new. That reasoning transfers, so we won't repeat it.
What doesn't transfer is what the search index has to work with. On Xiaohongshu the unit is a note with a body of text in it. On Douyin the unit is a video, and that changes the problem completely.
What the platform can actually read
This is the part that changes how you shoot. By most accounts of how the platform behaves, Douyin's search reaches past the caption: it is commonly described as reading text that appears on screen and transcribing what is spoken, so a keyword can be matched even when nobody typed it anywhere.
Which means the searchable surface of your video is larger than you think, and for an overseas operator it is usually empty:
- The caption is in English, or it's three emoji.
- The on-screen text is in English, or there is none.
- The audio is a music bed with no speech, or a guide talking in English.
- The account name is in Latin script, so nobody can type it even if they want to.
Four surfaces, no Chinese on any of them. The footage can be extraordinary and the video is still unreachable by a Chinese query, because there is nothing in it to match.
The words a traveller actually types
The words that matter are the ones a traveller would type, and those are almost never your company name. They're a place plus a problem: how many days, which month, is it suitable for parents, is there anyone who speaks Chinese, how do you get there from the nearest city.
So the caption should read like the answer to a question somebody asks out loud. "3 天 2 夜够不够看亚马逊?" earns its place in a way that "Discover the magic of the rainforest 🌿✨" cannot, because the first is a string somebody types and the second is a string nobody types.
Put the important words early. Captions get truncated, and the first line is the part that is read and the part most likely to be indexed cleanly.
Subtitles are an index, not an accessibility feature
Burn Chinese subtitles into the video. It does three things at once: it makes the video watchable with the sound off, it puts Chinese text on screen where the platform can extract it, and it removes the need for the viewer to understand your guide's accent.
One trap worth naming. Automatic captions transcribe the audio, so if your guide speaks English, the automatic caption is English. That's a correct transcription and a useless index. Chinese subtitles have to be added deliberately, from a translation of what was said, rather than left to the tool.
Where a Chinese voiceover is possible it's better still, because then the spoken layer carries the keywords too. But subtitles alone get you most of the way, and they don't require anybody to speak on camera.
Give yourself a name people can type
If your business only exists in Latin script, a viewer who wants to find you again has to remember an English string and switch keyboards to type it. Most won't. A short Chinese name, used consistently across the account name, the captions and the pinned comment, is what makes the second search possible — and the second search is where a brand starts to exist.
Consistency matters more than cleverness here. Three transliterations spread the mentions across three names, which is the same fragmentation we described in where Chinese reviews live.
Tag the place
Location is a query in its own right, and a specific place name is a narrower thing to compete on than a product category. Claiming and tagging the location connects your video to a surface that people browse deliberately, which we covered in Douyin POI pages.
What search optimisation cannot do
It only works if the query exists. If nobody is typing your city plus your category in Chinese, there is no search demand to capture, and no amount of caption craft creates it. Search is a way to be found by people already looking; it is not a way to make people look.
That's worth checking before you invest. Search the obvious phrases yourself, in Chinese, and see whether anything comes back and how good it is. If the results are thin, that can mean opportunity or absence, and the difference matters. Feed content and search content also aren't the same shape — a video built to stop a scroll is not necessarily the video that answers a question, and trying to do both in fifteen seconds usually does neither well.
Where CN1X fits
We take footage you already have and make it findable: captions written as the phrases a traveller types, Chinese subtitles added to the video, a Chinese name chosen once and used consistently, and the location tagged. The account is registered to your company and stays yours, along with the content and the data.
We can buy placement, and we still won't promise a ranking, because nobody can honestly promise one — paid reach and search ranking are different things, and this post is about the second. What we commit to is what gets published and how often. If you have video sitting unused from last season, tell us what you sell and we'll say what could be made findable from it.
More from the blog
Korean and Japanese Operators Selling into China
A short flight changes the customer. What long-haul advice gets wrong for operators in Korea and Japan, and which barriers you have already cleared.
What a Month of Running the Account Actually Looks Like
"What is the ongoing commitment" is really a question about calendars. Where the hours land, who they land on, and which parts cannot be batched.
What You Supply Before a Mini Program Build Starts
The quote is signed and then three weeks pass with nothing visible happening. Almost always the build is waiting on things only you can hand over.


