Overall Strategy

• Answers first
• Ops team at the core; product+engineering driven
• Creator‑centric; creation is king
• Zhihu positions as a deep‑content community

Product/Engineering Focus

• Treat moderators/answerers like community owners

Recommendation Algorithm Updates

• Cold‑start optimization
• Objective shift for ranking
• Community governance (reduce flamebait/opposition; ~2% comments removed)
• Health of the ecosystem; encourage professional verticals

Past: PV/UV and traffic were king. Now: depth and interactions matter more.

Hot List Experience

• Reduce confrontational topics
• Content ops strategy tweaks
• Product: remove “collapse”, add “featured”
• Question page ranking: suppress low quality; optimize TPR

True Positive Rate

True Positive Rate (TPR, recall/sensitivity) measures how well a classifier identifies positives: TPR = TP / (TP + FN). TPR ranking compares configurations/models by their TPR to choose better settings. In contexts like screening or fraud detection, high TPR is crucial.

User Communication

• Feedback loops

AI Content Stance

• Opinion extraction, similarity, quality, trust
• Plagiarism detection and enforcement
• Anti‑scraping (SEO shielding?)
• AIGC/Spam detection

Actions: restrict distribution; penalize malicious accounts. For Zhihu’s own “AI direct answers”: what sources were quoted to form the answer?

Author Experience

Good Questions

• Asking is somewhat de‑emphasized
• How are duplicates handled? Anti‑cheat for mass question spammers?

Notifications

• Improvements to reminders

Columns

• Upgrades, exposure, circulation, paid support
• What about Zhihu Daily?
• High SEO and content quality (reopen when?)

AI Search / Direct Answers

• TBD