On 6 August 2026 the China Advertising Association (CAA) held a second comment meeting in Beijing on the GEO group standard. Xinhuanet and People's Finance reported attendees from CCTV.com, Douyin Group, Alibaba Qwen, Sohu, 360 Zhijian, Weimob, Leo Digital and the National Advertising Institute; a submission draft would follow.
According to People's Finance (9 August 2026), CAA Secretary-General Huo Yan and Deputy Secretary-General Cui Yan attended the discussion in person. The Eastmoney summary (2026-08-09) adds Shengyan Intelligent, PureblueAI, Titanium AI, Oxygen Technology, Frost & Sullivan and Taihe Tai law firm among the platform, model-vendor, media, consulting and research representatives along the GEO industry chain.
GEO is moving from "everyone for themselves" to "there is a standard" — a constraint, and a window to occupy AI recommendation before the rules harden.
Where the standard stands
The CAA launched GEO standardization in early March 2026, leading the drafting of the GEO Trusted Information Dissemination and Information Ecosystem Governance standard — the first domestic group standard focused on trusted GEO dissemination (Xinhuanet, 2026-03-19). CAA president Zhang Guohua has said the work responds to service capabilities that are uneven and industry norms that are still missing in the fast-growing GEO sector (Xinhuanet, 2026-03-19). Drafters and drafting organizations were solicited after project approval (Sohu Finance, Baidu Baike).
| Date | Progress |
|---|---|
| Early March 2026 | CAA launches GEO standardization; drafting preparation begins (Xinhuanet, 2026-03-19) |
| 8 July 2026 | First public comment, industry and regulator feedback |
| July 2026 | Project approved, drafters solicited |
| 6 August 2026 | Second comment meeting in Beijing, heading to submission draft |
| TBA | Submission draft review and formal release |
Why a standard? CAA listed four GEO risks: rising information pollution and misleading content; uneven vendor quality and exaggerated promises; inconsistent measurement; non-compliant competition squeezing honest players (Xinhuanet). These hit answer quality and the trust assets brands pay for.
The stakes keep rising with usage. QuestMobile's Q1 2026 AI Application Insights (published 2026-04-21) put China's AI native apps at 440 million MAU, with Doubao at 345 million. By May 2026 that figure reached 499 million, up 85.4% year on year (QuestMobile H1 2026 report, published 2026-07-14). More AI answers mean more brand reputation risk — and more room for both compliant and non-compliant influence.
The red line: white-hat vs black-hat
The draft first draws that line, banning corpus poisoning, answer monopoly and prompt injection, and proposing "three-way separation" of brand knowledge bases — facts, opinions and marketing kept apart so AI citations stay reliable (Sohu Finance).
| Dimension | White-hat (compliant) | Black-hat (banned) |
|---|---|---|
| Sources | Verifiable facts, multi-source cross-check | Homogeneous bulk, corpus poisoning |
| Content | Facts, opinions, marketing separated | Answer monopoly, manipulated answers |
| Methods | Structured content, knowledge bases | Prompt injection, fake volume/reviews |
| Outcome | Long-term trust assets | Demotion, wipe, reputation damage |
This aligns with Doubao's 16 August update: "no single-source proof," bulk AI advertorials cleared (see GEO after Doubao's update). Platform rules and industry standards are tightening together; black-hat room shrinks. The gap between tactics is measurable: the State of AI Search 2026 study reports an average brand mention rate of just 17.2% across AI answers — most brands are not cited at all, and manipulation mainly raises the chance of being filtered.
Four steps to a compliant layout before the standard lands
1. Three-way knowledge base. Fact layer (profile, credentials, specs, process), opinion layer (insights, white papers, case reviews), marketing layer (promos, slogans) — presented separately so marketing does not pollute fact retrieval. Core of the draft and of how engines score trust (see Brand AI mention rate).
2. Multi-source matrix. QuestMobile Q1 2026 put Doubao above 345 million MAU; the June list still showed Doubao first at over 380 million (Sina Tech, 2026-07-14). One official site cannot support cross-check; cover encyclopedia, trade media, third-party reviews and official social with the same facts (see GEO intro).
3. Unify measurement. The standard calls out inconsistent monitoring. With average mention rates as low as 17.2% (State of AI Search 2026), a brand that tracks mention, position and citation sources across engines with one definition of ROI can see clearly whether its source matrix is working.
4. Pick a white-hat vendor. After the standard, the professional bar rises and black-hat shops clear out. Shanghai Zheming offers GEO services from knowledge-base cleanup to measurement.
FAQ
Is the group standard mandatory? It is voluntary, not statute. It is a reference for self-regulation and platform governance. Doubao, Yuanbao and Qwen rules are already moving toward it — earlier compliance is more agency.
What counts as black-hat GEO? The draft bans corpus poisoning, answer monopoly, prompt injection and bulk homogeneous content that interferes with inclusion.
Is the standard good for SMEs without black-hat resources? Yes. Black-hat depends on spend and manipulation; SMEs cannot compete there. The standard returns competition to content quality and source trust.
What should we do now? Three steps: map the knowledge base to three-way separation, fill multi-source coverage, stand up AI-search measurement; then adjust when the submission draft publishes.
How do we assess current compliance of our AI presence? Search core terms on Doubao, Yuanbao, Qwen and DeepSeek; log mention, which sources were cited, and accuracy. Misattribution and distortion usually mean a source gap.
Related reading
Comment on the GEO group standard marks a norms phase: competition shifts from grabbing inclusion and volume to trust and compliance. Firms that finish knowledge-base governance and multi-source building first will have a lead when the standard lands. For a free AI visibility diagnosis: +86 18917757529, jaysun@widesight.cn.
This article was written by Zheming Digital Communication Research Institute. Data updated to 2026. Standard status follows CAA and the national group-standards platform; data cited from QuestMobile Research Institute public reports (2026-04-21, 2026-07-14) and the State of AI Search 2026 study.