Xiaohongshu AI search results can vary by user, time, available content and product changes. Brands should use a fixed question set to observe trends and separate mentions from answer accuracy, account health and business value. Unavailable internal platform data must not be presented as a deterministic ranking metric.
Create 30–50 core questions
Source questions from search suggestions, Ask, comments, support and enquiries, spanning category understanding, scenario selection, comparison, price or risk and post-purchase issues. Record audience, expected facts, supporting posts and refresh frequency. Keep 70% fixed month to month and use the rest for emerging demand.
Four score groups
- Coverage: whether the brand or its content appears for relevant—not unrelated—questions
- Accuracy: correctness of brand, service, pricing conditions, audience fit and risks
- Evidence: whether the answer can be traced to clear, authentic and current content
- Health: originality, disclosure, genuine comment handling and absence of repetitive publishing
Move from diagnosis to action
If coverage is low but accuracy is high, fill unanswered questions. If coverage is high but facts are wrong, govern brand information and update prominent content first. If saves are strong but enquiries weak, review the profile, service explanation and CTA. If engagement changes abruptly, investigate media spend, campaigns or platform changes before duplicating content. Change one variable group at a time and retain an edit log.
What a monthly report should include
Include test dates and conditions, question coverage, material errors, rising and declining content, new user questions, account risks and next month’s owners and updates. Distinguish platform-visible data, team observations and inference so correlation is not reported as causation.
Is daily testing necessary?
Not for most brands. Weekly spot checks and a fixed monthly test are usually more comparable; increase frequency temporarily around campaigns or major product changes.