Xiaohongshu Product Seeding Prompts for Authentic 种草

13 prompts for native-reading Xiaohongshu 种草 posts — first-person, scenario-based, honest pros/cons — plus the 2026 disclosure and AI-label rules.

Xiaohongshu (小红书) seeding fails the moment a post reads like an ad: uniform 5-star tone, a feature list instead of a scenario, no real “I tried it, and here’s the catch” moment. The 2026 algorithm makes this worse, not better. The platform now runs a machine-plus-human dual review that screens for hidden soft ads, absolute-language banned words (违禁词 like 最/第一/顶级), and unlabeled AI text — and it limits distribution on anything that trips those filters. Posts under ~600 characters or below 60% originality also get throttled.

These 13 prompts force the four things that survive that review: first-person voice, one concrete scenario, 1-2 honest downsides, and a pro/con frame that builds trust even on a paid post. They are written for a current model (GPT-5.5, Claude Sonnet 4.6, or Gemini 3.1 Pro — all free-tier capable for short copy). Treat the AI output as a first draft you personalize, then add your own photos and the required labels before publishing.

TL;DR

  • Seeding dies on a perfect 5-star tone. Every prompt below bakes in 1-2 honest downsides and a who-it’s-not-for line.
  • Xiaohongshu’s machine + human review (tightened across 2026) throttles hidden soft ads, absolute words (违禁词), and unlabeled AI content — labeling AI output has been mandatory since the platform’s 12 Feb 2026 announcement.
  • Aim for 600+ characters and 60%+ original material (real photos, your own usage). Notes around 800-1,200 characters tend to rank best in search.
  • Paid collabs must route through 蒲公英 (Pugongying), which stamps a 赞助 / 品牌合作 label. Bury that label and you risk a limit or a ban.
  • Use a current model (GPT-5.5 / Claude Sonnet 4.6 / Gemini 3.1 Pro) for the draft; rewrite in your own voice so it reads native, not generated.

Best for

  • Brand-collab 种草 posts routed through 蒲公英
  • Founder self-promo posts for a DTC product
  • KOC seeding for new DTC brands
  • New-product launch 种草
  • Niche-category seeding (beauty, lifestyle, food, tech)

How to use these prompts

Replace each [bracketed] placeholder with your own detail, then paste into ChatGPT, Claude, or Gemini. Two rules keep the output native rather than generated:

  1. Feed it your specifics. The more real detail you give (what you tried before, the exact moment it clicked, the price you paid), the less the draft reads like template copy. Vague input produces vague, ad-shaped output.
  2. Rewrite, then label. Edit the draft into your own voice and swap in real photos. If you keep AI-generated phrasing or visuals, add Xiaohongshu’s AI label before publishing — unlabeled AI content is down-ranked as of February 2026.

1. Day-in-life scenario seeding

Write a Xiaohongshu 种草 post about [product]. Format: a real day-in-life
scenario where [product] solves one specific moment. ~300 characters.
Include: setting, what I used before, why I switched, 1 honest downside,
who would love this. First-person, casual, no absolute words like "best".

2. First-impression review

Write a "first 7 days using [product]" review for 小红书, ~250 characters.
Format: day 1 expectations, day 3 surprise, day 7 verdict. Include 2 things
I liked and 1 thing I am still unsure about. Conversational, not promotional.

3. Compare-against-staple seeding

I have used [staple product] for [time]. Now trying [new product]. Write a
~280-character compare-and-contrast 种草 post. Be honest about which wins on
which dimension. Do NOT give the new one a clean sweep of 5 stars.

4. Problem-specific seeding

I had this specific problem: [paste problem]. Write a ~280-character post on
how [product] solved it. Include the failed attempts before it, the moment
it clicked, and 2 caveats. Concrete and specific, not generic.

5. Cheap-vs-luxury seeding

I tried both the cheap [category] version ([price]) and the luxury one
([price]). Write a ~300-character 种草 post comparing them. Honest verdict on
whether the extra cost is worth it for [persona], and when the cheap one is enough.

6. Skin-type / body-type / fit specific

Write a ~250-character seeding post for [product] specifically for [audience —
e.g. dry skin / petite frame / sensitive scalp]. State exactly what to do and
what to avoid for this audience. Mark one situation where it would NOT suit them.

7. Routine-integration post

Write a ~280-character post on how [product] fits into my existing [routine —
skincare / morning / cooking]. List the order I use it in, what NOT to combine
it with, and the timing trick that maximized results.

8. Honest-warning post (after the honeymoon)

I have used [product] for 60 days. Write a ~250-character honest follow-up:
3 things that lived up to the hype, 2 things that did not, 1 thing I wish I
had known before buying. Authentic over glowing.

9. Title generator (Xiaohongshu specific)

For my 种草 post about [product], generate 8 title variants in Chinese. Use
formats that work on 小红书: question titles, numbered ("3 个 X 我不会再回头"),
emotional, and "before vs after" hooks. Keep to ~16 Chinese characters where
possible. Avoid absolute words (最/第一/顶级).

10. Hashtag / topic-tag set

For my [category] 种草 post, suggest 12 relevant 话题 / hashtags ordered by
reach. Mix: niche-specific, persona, scenario, and evergreen. Avoid generic
spam tags and avoid banned absolute terms.

11. Pro/con frame for credibility

Write a ~280-character 种草 post that opens with "I am not going to lie about
this — here are the real pros and the real cons." Then 3 pros and 2 cons. End
with who this is and is not for. The goal is to keep trust on a paid post.

12. Repurpose to TikTok / Reels

Below is my 小红书 种草 post. Repurpose it into: (a) a 60-second TikTok / Reel
script, (b) an Instagram caption with 4 carousel slide notes, (c) a 6-post X
thread. Preserve the honest tone and the downsides.

[paste post]

13. Brand-collab disclosure version

Write a brand-collab 种草 post that includes the required #广告 / #合作
disclosure naturally near the top, not buried at the end. ~280 characters.
Keep it authentic: include what I would say even if I were not paid, and name
one place the brand brief did NOT match my real experience.

2026 Xiaohongshu rules these prompts respect

The platform tightened enforcement across 2026, and AI-assisted copy is a named target of the review system. Build these into your workflow so a good post is not throttled on a technicality.

Rule (as of June 2026)What it means for a seeding post
AI-label requirement (since 12 Feb 2026)Add Xiaohongshu’s AI / synthesized-content label if you keep generated text or images. Unlabeled AI content is down-ranked.
Dual review (machine + human)Avoid hidden soft ads, banned absolute words (违禁词), and traffic-baiting links. Use “亲测有用” / “适合 XX 人群”, not “最好用”.
Originality ≥ 60%Real photos, your own experience, honest before/after. Polished brand copy alone gets throttled.
Length floor ~600 charactersVery short notes are limited; ~800-1,200 characters tend to rank best in search.
Paid-collab reporting via 蒲公英Route paid posts through Pugongying so the 赞助 / 品牌合作 label appears. Burying disclosure risks a limit or ban.

Two notes on the numbers: Xiaohongshu reportedly removed roughly 600,000 low-quality AI notes in the first half of 2025, which is why the labeling rule arrived in February 2026; treat all of the above thresholds as the platform’s stated guidance, not a guaranteed formula.

Which model to draft with

All three current consumer models handle short Chinese seeding copy well on free tiers; the difference is mostly voice and context.

Model (June 2026)Free tierBest for here
GPT-5.5Yes (tight limits)Punchy hooks and title variants; default in ChatGPT since ~Apr 2026
Claude Sonnet 4.6Yes (limited)Longer, more natural first-person narration; least “ad-shaped”
Gemini 3.1 ProVia Google AI free tierGood Chinese idiom; useful if you also pull in Google data

For a long honesty-driven post (prompt 8 or 11), Claude tends to keep a calmer, less salesy register. For snappy titles (prompt 9), GPT-5.5 is fast and varied. Whichever you use, the draft is a starting point — the original 60% has to come from you.

Common mistakes

  • A perfect 5-star tone with no downside — readers smell the ad and bounce, and the review system flags hidden soft ads.
  • Absolute words (最 / 第一 / 顶级) — banned terms that can throttle or remove the note; swap for “亲测” or “适合 XX 人群”.
  • No real scenario — a feature list with no “I tried this and…” moment fails both readers and the originality check.
  • Publishing AI text or images without the AI label — automatic down-ranking since February 2026.
  • Hiding or burying the #广告 / #合作 disclosure, or skipping 蒲公英 on a paid post — platform penalty plus reader distrust.
  • The same caption template recycled across unrelated products — kills originality scoring and reads like a bot.

FAQ

Will Xiaohongshu penalize a post I wrote with AI? Not for using AI, but for not labeling it. Since the platform’s 12 February 2026 announcement, AI-generated or synthesized text and images must carry the AI label; unlabeled AI content gets distribution limits. Rewrite the draft in your own voice and add real photos so most of the note is genuinely yours (the platform looks for 60%+ originality), and label anything that stays AI-generated.

How long should a seeding note be? Clear the ~600-character floor — shorter notes get throttled. Notes around 800-1,200 characters tend to perform best in search, so the ~250-300 character drafts these prompts produce are a core you expand with your own detail, not a finished post.

What words get a seeding post limited? Absolute-language banned terms (违禁词) such as 最 / 第一 / 顶级 / 国家级, plus medical or efficacy overclaims. The 2026 dual review (machine first pass, then human) screens for them. Replace with hedged, testable phrasing like “亲测有用” or “适合 XX 人群”.

Do I have to disclose a paid collaboration? Yes. Paid brand collabs must run through 蒲公英 (Pugongying), which applies the 赞助 / 品牌合作 label, and you should keep a #广告 / #合作 tag visible near the top. Hiding it risks a traffic limit or a ban, and readers notice anyway.

Which AI model is best for Xiaohongshu copy? Any current model works on its free tier: Claude Sonnet 4.6 for natural, less salesy first-person narration; GPT-5.5 for fast title and hook variants; Gemini 3.1 Pro for solid Chinese idiom. The model only drafts — the original detail and the honest downsides have to come from you.

Tags: #Prompt #Social media #Xiaohongshu