Xiaohongshu search: how to get found on Red, not just post to it
Short version: on Xiaohongshu, posting is not the same as being found. Red is a search engine wearing a feed's clothes — nearly seven in ten of its monthly users search on it, and a third open the app and go straight to the search bar. If your notes aren't built to surface when someone types the question your product answers, you're paying to talk to the algorithm, not to buyers. Getting found means treating each note as a search result: the right keywords in the right places, and the engagement signals Red actually weights — saves over likes, every time.
Here's the thing most foreign brands miss about Red. They pour a quarter's budget into a burst of pretty notes, watch the impressions tick up for a week, and then wonder why the account goes flat. What they built was a feed campaign. What Red rewards, and what compounds, is search. A note that ranks for "sensitive-skin sunscreen" keeps pulling in buyers months after you posted it — because on Red, people don't just scroll, they look things up. Nearly 70% of monthly users run searches, and since late 2024 the platform has been handling on the order of 600 million searches a day. Around 200 million users a month come specifically to research a purchase. That's not a feed. That's Google for your category, and almost nobody is optimizing for it.
Why search beats the feed on Red
Feed traffic is a lottery you re-enter every time you post. It spikes, then decays, and the next note starts from zero. Search traffic is an asset. Someone typing "postpartum hair loss shampoo" into Red is further down the funnel than anyone the feed served — they've named the problem and they're shopping for the answer. The agencies that run these accounts all report the same thing: search-sourced visits convert meaningfully better than feed-sourced ones. I won't put a precise multiple on it because the numbers people quote vary and few are auditable, but the direction isn't in dispute, and it matches basic intent logic. Higher intent, lower volume, far better economics.
So the goal isn't "go viral." Viral is nice and mostly unrepeatable. The goal is to own the searches your buyer actually types — the boring, specific, high-intent ones — and to keep owning them.
Write for the query, not for yourself
Red searches are conversational and specific in a way Baidu searches aren't. People don't type "lipstick." They type "lipstick that doesn't transfer to a mask" or "matte lipstick for morning-brown lip tone." That specificity is the opportunity: the broad terms are a bloodbath of established brands, but the long tail is wide open, and it's where purchase intent actually lives. A useful working split is roughly 70% long-tail, problem-shaped keywords and 30% broad category terms — the long tail is what you can realistically rank for, the broad terms are a bonus when a note over-performs.
Do the keyword work before you write, not after. Type your seed terms into Red's own search bar and read the autocomplete suggestions and the "related searches" it offers — that's the platform handing you the exact phrasing real users use, for free. Build your note around one of those phrases instead of retrofitting a caption onto a photo you already shot.
Where the keywords go
Red's ranking isn't a mystery box. It leans on a few things you can influence: how well the note matches the query, how much the audience engages with it, how much authority your account has built, and whether the topic is trending right now. Keyword placement feeds directly into the first of those. Not all positions carry equal weight — the title and the opening lines matter far more than a pile of hashtags at the bottom.
| Placement | Weight | How to use it |
|---|---|---|
| Title | Highest | Lead with the exact query phrase; keep it tight (Red truncates long titles — think ~20 characters in Chinese) |
| First ~100 characters of the body | High | Restate the keyword naturally in the opening line, where both the reader and the algorithm look first |
| Text baked into the images | Medium-high | Red reads on-image text; a cover with the keyword doubles as your thumbnail hook |
| Hashtags | Medium | A handful of relevant tags — mix one or two broad with several niche; don't stuff 30 |
| Body + closing line | Supporting | Repeat the term once or twice more, plus close variants — natural density, not keyword soup |
The failure mode here is over-optimization. Cram the same phrase in eight times and the note reads like spam to humans and gets throttled by the algorithm, which is tuned to demote exactly that. Write for a person who's genuinely looking for help; place the keyword where it naturally belongs; stop.
Saves are the signal, not likes
This is the part most Western marketers get backwards. On Red, the engagement hierarchy that moves search ranking runs roughly saves > shares > comments > likes. A like is a reflex. A save (收藏) is a declared intention — "I'm coming back to this when I buy" — and Red treats it as the strongest vote that a note is genuinely useful. That single fact should reshape how you write. Stop making posts that earn a thumb-tap and start making posts worth bookmarking: the comparison table, the step-by-step routine, the "which one for which skin type" breakdown, the honest list of who it's not for. Reference content gets saved. Pretty content gets liked and forgotten.
Comments matter too, and they're gameable in the honest way: end the note with a real question, and answer every reply. Early engagement in the first few hours weighs heavily, so posting when your audience is actually awake beats posting when it's convenient for your team's timezone.
Account authority compounds — or leaks
Red also weighs the account behind the note. A profile that posts consistently in one lane, keeps its content ratings clean, and avoids the obvious tripwires (external contact info in captions, hard-sell language, anything that reads as an unmarked ad) accrues weight, and its new notes start ranking faster. An account that spams, buys obvious fake engagement, or lurches between unrelated topics leaks that weight and gets a colder start every time. Consistency isn't a virtue here, it's a ranking input. This is also why a coordinated set of KOC and mid-tier creator accounts outperforms one brand account shouting alone — you're building searchable coverage across many trusted profiles, not one.
A simple way to start
- Pick five long-tail queries your buyer actually types. Pull the exact phrasing from Red's autocomplete.
- Write one genuinely useful, save-worthy note per query — reference content, not a brochure.
- Put the query in the title, the first line, and the cover image. Once more in the body. Stop there.
- Post when your audience is awake; answer every comment in the first few hours.
- After two weeks, search those queries yourself and see where you rank. Double down on what moved.
Bottom line
Most brands lose on Xiaohongshu because they treat it like a feed to fill instead of a search engine to rank in. Flip that. Find the specific, high-intent queries your buyer types, write reference-grade notes built around them, place the keyword where it counts, and optimize for saves rather than likes. Do that consistently and you stop renting attention one post at a time and start owning the searches that lead to a sale. It's slower than a viral hit and worth far more.
If you want a read on which Red searches are actually worth owning in your category — and whether Red is even the right first platform for you — that's the work I do; reach out. And if you're still deciding where to build first, start with Red versus Douyin: which to build on first.