TL;DR
Xiaohongshu titles display up to 20 Chinese characters, and the home-feed waterfall wraps after roughly 10, so the first 10 characters do almost all the work. Feed your AI model one sentence of topic, the reader’s exact pain, and 3-5 current high-performers from your niche, then ask for 15 titles grouped by hook style. Put your search keyword inside the first 10 characters, keep one specific noun or number per title, and pick the winner yourself. This page has the copy-ready prompt and the rules to check the output against.
The task
You publish on Xiaohongshu (RedNote) and your titles do not earn taps. The feed is dense, and competing notes use emoji, numbers, contrast, and pain-naming. You want a batch of title options that match your content and your niche’s current style, not the same five formulas every account on the platform recycles.
Two platform facts shape every good title:
- The hard cap is 20 characters. Past that, the title is truncated. The home-feed waterfall typically wraps after about 10 characters, so the first line is what people actually scan.
- The first 10 characters drive search. Xiaohongshu doubles as a search engine, and your front-loaded keyword influences how the note ranks for that query. A clever hook that buries the keyword at character 15 costs you discovery.
When AI is the right tool
- You are publishing 2+ times a week and titling is the slowest step.
- You want to A/B 3-5 titles against the same note.
- You know your audience pain but struggle to compress it into 20 characters with the keyword up front.
- You want to break out of repeating your own past title patterns.
When not to rely on AI alone
- For slang or memes that emerged in the past month — model training data lags, and Xiaohongshu trends move weekly.
- For high-stakes brand notes where one off-brand title costs trust.
- In a sub-niche the model clearly does not know (very local communities, narrow professional jargon).
Which AI model to use
Title generation is a Chinese-language hook task, so the model’s grip on idiomatic Mandarin matters more than its benchmark score. As of June 2026:
| Model | Strength for Xiaohongshu titles | Cost |
|---|---|---|
| DeepSeek / Kimi (Chinese-native) | Best feel for current Mandarin phrasing and platform slang; free web tiers | Free / low |
| GPT-5.5 | Strong bilingual hooks, reliable instruction-following on the 20-char rule | ChatGPT Plus $20/mo |
| Gemini 3.1 Pro | Good Chinese output, 1M-token context for pasting many reference titles | Google AI Pro $19.99/mo |
| Claude Sonnet 4.6 | Clean structure and rationale; slightly more formal Mandarin register | Claude Pro $20/mo |
For pure Chinese hook phrasing, a Chinese-native model (DeepSeek or Kimi) often lands closest to how creators actually write. If you already pay for ChatGPT Plus or Google AI Pro, those handle the task well too — the bottleneck is your input, not the model.
What to feed the AI
- The note topic in one sentence.
- The reader’s specific pain or curiosity (“they think X is hard but it is not”).
- The search keyword you want to rank for (this goes in the first 10 characters).
- The hook style you want: contrast, listicle, before/after, secret reveal, warning.
- Niche reference: 3-5 top-performing titles from accounts in your niche this week.
- Constraints: 20-character cap, emoji policy, claim words to avoid.
Copy-ready prompt
Generate 15 Xiaohongshu (RedNote) note titles in Chinese.
Topic: "[topic]"
Reader pain: [what they currently struggle with]
Reader desire: [what outcome they want]
Search keyword to rank for: [keyword]
Hook styles to mix: curiosity, conflict, list, instruction, before/after, warning.
Reference high-performing titles in this niche right now:
- [title 1]
- [title 2]
- [title 3]
Constraints:
- 20 Chinese characters MAX per title (hard platform cap).
- Put the search keyword inside the first 10 characters.
- Use emoji only where it lifts the hook, not on every title.
- No empty buzzwords (e.g., filler superlatives that just mean "amazing").
- Each title must contain ONE specific noun or number.
Output:
- 15 titles, grouped by hook style.
- A 1-line note for the top 3 explaining why each should work.Recommended output structure
Group the 15 titles by hook style (5 styles, 3 titles each) and mark the 3 strongest with a short rationale. This makes the list easy to pick from and, over time, shows you which styles your niche rewards. Notes that pair a specific number or timeframe with the hook tend to lift read-through, so favor titles that name a count, a duration, or a price.
How to check the output
- Count characters: each title must be 20 or fewer. Tools cut at the cap.
- Is the keyword inside the first 10 characters? If not, rewrite the front.
- Read each title in 1.5 seconds — would you tap?
- Does each title contain a specific noun, number, or named pain?
- Do they sound like your account, or like a generic agency wrote them?
- Compare with this week’s top 5 notes in your niche — is yours in the same energy?
Common mistakes
- Buzzword stuffing. A title made entirely of empty filler superlatives (“absolutely amazing must-see gorgeous tutorial for sisters”) earns nothing and burns characters.
- Burying the keyword. A great hook with the search term at character 15 wastes your ranking shot. Front-load it.
- Generic pain. “How to be productive” is dead; “how I stopped opening Xiaohongshu before 9am” works.
- Letting AI pick. Models lean toward safe, middle-of-the-road titles. Generate 15, but you choose the one to ship.
Next steps to keep improving
Track which titles earn the highest tap-through rate. After about 20 notes you will see your account has 2-3 hook styles that consistently outperform. Bake those into your prompt as the default mix and drop the rest. Re-feed fresh reference titles every couple of weeks, because the styles that win shift with the platform.
For deeper background on the algorithm and what the feed rewards, see Xiaohongshu’s own creator guidelines and any current third-party teardown of the search ranking.
FAQ
What is the title character limit on Xiaohongshu? 20 Chinese characters. Anything longer is truncated, and the home-feed waterfall usually wraps after about 10, so the first 10 characters carry the hook.
Where should the keyword go? Inside the first 10 characters. Xiaohongshu is a search engine as much as a feed, and a front-loaded keyword helps the note rank for that query.
Should I use emoji? Selectively. One emoji can lift a hook; notes that overuse emoji read as bait and waste your 20-character budget.
How many titles should I generate? Generate 15, keep 3, ship 1, and save 2 for an A/B test on similar notes.
Will AI know current platform trends? Only partially — training data lags by months. Always paste 3-5 current high-performers from your niche so the model matches this week’s energy.
Which AI model is best for Chinese titles? A Chinese-native model like DeepSeek or Kimi usually nails the phrasing; GPT-5.5, Gemini 3.1 Pro, and Claude Sonnet 4.6 all handle it well if you already subscribe.