📅 1. July 15: The Day AI Marketing Changed Forever
On July 15, 2026, two apparently unrelated events occurred on the same day — and together they redraw the boundaries of AI marketing in China.
At midnight, Douyin e-commerce's revised service fee rules took effect, pricing Doubao-sourced traffic for the first time. Hours later, Doubao itself shut down its user-created AI agent feature — along with Alibaba's Tongyi Qianwen and Tencent's Yuanbao — purging a combined ecosystem of over 8 million AI agents from their platforms.
These were not product optimizations. They were not A/B tests. They were compliance actions triggered by the same regulatory deadline: July 15, 2026, the effective date of the Artificial Intelligence Personified Interaction Service Management Interim Measures, jointly issued by five national ministries including the Cyberspace Administration of China (CAC) and the National Development and Reform Commission (NDRC) on April 10, 2026.
| Timeline | Event |
|---|---|
| April 10, 2026 | Five ministries issue the regulation, 3-month transition period begins |
| June 2026 | Tencent Yuanbao quietly closes agent creation entry |
| July 15, 2026 | Regulation takes effect; Doubao (8M+ agents) + Tongyi Qianwen shut down simultaneously |
The message is unambiguous: China's three largest AI platforms collectively decided that compliance through agent-by-agent auditing was impossible at scale. The only viable option was deletion.
🚫 2. Four Red Lines: What the Regulation Actually Bans
The regulation establishes four prohibitions that directly impact how brands can use AI for marketing:
| Prohibition | What Is Banned | Marketing Practices Eliminated |
|---|---|---|
| ① No AI impersonation of humans | AI agents cannot present themselves as real people | Virtual sales consultants, AI customer service posing as human |
| ② No inducing emotional dependency | AI cannot build emotional bonds to drive behavior | AI companions (情感搭子), emotional loyalty marketing |
| ③ No blurring human-AI boundaries | Users must always know they are interacting with AI | Unlabeled AI interactions, "AI passes as human" scenarios |
| ④ No personified marketing language | AI cannot use emotionally manipulative sales scripts | "AI trust-building" techniques,套路话术 conversion |
Beyond these prohibitions, the regulation requires: mandatory AI identity disclosure, scenario-specific service boundaries, upgraded content moderation systems, and user rights protection mechanisms.
The regulation specifically targets the gray zone that had developed over three years: AI agents that mimicked human tone, personality, and communication style to reduce user defenses and drive conversions. Industry estimates suggest Doubao alone had accumulated over 8 million user-created agents — from virtual advisors to AI sales reps to emotional companions — many of which crossed at least one regulatory red line.
🗑️ 3. The Platform Response: 8 Million Agents Deleted
Each of the three major platforms responded differently — but all reached the same conclusion about the central issue:
| Platform | Action | Scale | Migration |
|---|---|---|---|
| Doubao (ByteDance) | Shut down all user-created agents | 8M+ agents | Migrated to Maoxiang (猫箱) — independent, niche, risk-isolated |
| Tongyi Qianwen (Alibaba) | Shut down all agents | Undisclosed | None — no migration, no transition |
| Yuanbao (Tencent) | Closed agent creation entry | Preemptive | Entry removed before regulation took effect |
ByteDance's approach is the most instructive. Doubao is a billion-user super-app that cannot afford compliance risk from 8 million uncontrolled agents. The solution: quarantine the personification capability into a separate, niche app (猫箱) where scale is limited, risk is contained, and regulatory scrutiny is manageable. The core Doubao product remains clean — focused on utility, not personification.
Alibaba's approach is the most definitive. Tongyi Qianwen is positioned as a tool for e-commerce and life services, not emotional companionship. When compliance costs became clear, Alibaba simply amputated the functionality with no replacement. The signal: personified AI marketing is not a core Alibaba strategy.
Tencent's approach is the most cautious. Yuanbao closed its agent creation entry before the regulation even took effect, choosing to stay ahead of the compliance curve rather than react to it.
The combined message from the industry's three largest players: the era of personified AI marketing is over.
🔄 4. AI Marketing's Strategic Shift: Personification → Content Authority
The regulation creates a clear divergence between two AI marketing strategies:
| Strategy | Before July 15 | After July 15 |
|---|---|---|
| Personified AI agents | Growing channel, low barrier to entry | Illegal at scale, platforms are deleting |
| GEO (Generative Engine Optimization) | Emerging discipline | The only compliant AI marketing channel |
The reason GEO survives while personified agents are eliminated is structural, not tactical. GEO operates on a fundamentally different premise:
- Personified agents: AI talks to users as if it were a person → trusts is built through conversational mimicry → conversion happens through emotional manipulation → inherently crosses all four regulatory red lines
- GEO: Brands publish authoritative content → AI search engines index and cite it → users see brands cited as credible sources in AI-generated answers → trust is built through demonstrated expertise, not emotional manipulation
The regulatory logic is clear. Personified AI marketing exploits information asymmetry — users cannot reliably distinguish AI from human interaction. GEO operates on information transparency — users see brands because AI has determined their content is trustworthy, not because a persuasive agent has simulated a relationship.
For brands that have invested in personified AI marketing — virtual sales agents, AI companions, emotional marketing chatbots — the transition must begin immediately. Those agents will be deleted or become non-compliant.
⚡ 5. What Brands Must Do Now
| Priority | Action | Rationale |
|---|---|---|
| 🔴 Critical | Audit all current AI marketing touchpoints for personification risk | Identify and retire any AI agents, chatbots, or interactions that simulate human identity or build emotional dependency |
| 🔴 Critical | Remove any unlabeled AI interactions — every AI touchpoint must disclose its AI identity | Non-disclosure is a direct regulatory violation |
| 🟡 High | Shift AI marketing budget from personified agents to GEO (content authority building) | GEO is structurally compliant — it doesn't simulate humans, it builds authoritative content that AI cites |
| 🟡 High | Build a content authority strategy: knowledge graphs, structured FAQs, authoritative industry content | AI models cite content that is complete, consistent, and corroborated across multiple sources |
| 🟢 Medium | Monitor competitor migration — brands that move fastest to GEO will capture AI citation share before competitors arrive | First-mover advantage in AI citation is compounding |
📋 6. Key Takeaways
- July 15, 2026 is the AI marketing compliance watershed. Three major platforms deleted over 8 million agents on the same day — the effective date of China's first AI personification regulation.
- Four prohibitions define the new boundaries. No AI impersonation, no emotional dependency, no blurred human-AI lines, no manipulative sales scripts.
- Personified AI marketing is dead at scale. ByteDance, Alibaba, and Tencent independently concluded that compliance is impossible without deletion.
- GEO is now the only structurally compliant AI marketing channel. It does not simulate humans, exploit emotional dependency, or blur identity. It builds content authority that AI search engines cite.
- First-mover advantage in GEO is accelerating. As competitors retire their personified agents, brands that invest in authoritative content now will dominate AI citation share.
- The compliance audit deadline is now. Any brand still running unlabeled, emotionally manipulative, or human-impersonating AI interactions is operating in violation of law.