China's GEO Rulebook Tripled in a Week

In our September 11 analysis of China's GEO compliance reckoning, the story was enforcement: which agencies got filtered out once the August rules took effect. Four weeks later, the story has moved upstream.

In the space of a few days at the end of September, three separate standards bodies published or activated GEO frameworks. The market is no longer waiting for a single national rule. It is being governed by a stack — and the stack is growing faster than most brands outside China are tracking.

🗓️ A week of standard-setting

Two observations follow immediately. First, the national standard is still in development — so provinces and industry associations are filling the gap themselves. Second, there is no single rulebook. There are several, from several bodies, with overlapping scope.

📐 The five China Advertising Association standards

The CAA framework is the broadest in coverage. It spans five areas:

1. Terminology and general provisions — a shared vocabulary, which matters more than it sounds when four bodies are writing rules.
2. Provider assessment — how to evaluate a GEO service provider.
3. Pricing assessment and measurement — what GEO work costs and how it is measured.
4. Trusted corpus — what content is acceptable as source material.
5. Service-process compliance and safety — how engagements must be run.

The drafting process involved 41 participating organisations and 75 industry experts, processing 174 rounds of industry feedback. That level of consultation is unusual for a young category, and it signals that the standards are intended to be operational rather than aspirational.

📏 The Shandong specification: the operational one

The Shandong document, effective October 1, is the most concrete of the three. It defines Brand GEO as:

systematic optimisation targeting generative AI's semantic parsing logic and source-trust rules — building a brand knowledge system, deploying authoritative sources in tiers, and refining content expression — to improve the visibility, accuracy and positivity of brand information in generative AI output.

Three requirements in it will shape how work gets done:

1. Official approved brand information is the sole factual baseline. Not your marketing copy, not a partner's presentation deck, not a translated English page — the approved source of record.
2. AI-generated content must pass both human fact-verification and a compliance review before it can be distributed. Two checkpoints, both human, before publication.
3. A three-tier source system: a core source layer, an authoritative expansion layer, and a supplementary dissemination layer. Each tier carries different weight.

Its scope also extends well past commercial marketing — it covers product, corporate, industrial-cluster, regional, city, tourism, service and government public brands.

🔍 What these standards are — and are not

This distinction matters and is frequently blurred in coverage.

They are group standards, not national law. Group standards (团体标准) are voluntary frameworks issued by associations and provincial bodies. Nothing here is enforceable the way the August specification's prohibitions on corpus poisoning are.

But they are becoming gate conditions. Industry reporting indicates that Chinese buyers are already writing "has passed third-party trustworthiness evaluation" and "retains complete operation logs" into tender documents. A voluntary standard that appears in procurement requirements behaves like a mandatory one.

And the rulebook is not yet settled. With CAICT, the China Advertising Association, a provincial association, the China General Chamber of Commerce, and the Xinhua-backed August specification all active, China is in a period of competing standard-setting. Expect consolidation, and expect the requirements to move again.

🌏 Why the bar rises for overseas brands

Each of the Shandong specification's three requirements maps onto a capability that foreign companies typically do not have in China:

This is why the standards shift the problem rather than solving it. A brand that has been running translated English content through a third-party publisher now has to demonstrate a Chinese source of record, a documented verification process and evidence of tiered source placement. Those are operating capabilities, not content outputs.

The direction of travel is consistent with what we documented on September 10 in ERNIE's citation behaviour: the sources that carry weight in Chinese AI answers — Baidu Baike, Baijiahao, Baidu Zhidao — are exactly the surfaces where tier, provenance and traceability are now being formalised.

🧭 What to do, and where BPP fits

Practically:

1. Establish what your Chinese source of record is. Before any campaign, identify the authoritative Chinese-language version of your brand facts — and who owns it.
2. Build the verification step into your workflow. Two human checkpoints before publication is now the documented expectation, not an overhead you can skip.
3. Map your sources against the three tiers. Know which of your placements are core, which are authoritative expansion, and which are supplementary.
4. Expect the requirements to change. Standards are still being written by multiple bodies. Build processes that can absorb new criteria rather than optimising for one document.
5. Treat tender-level criteria as the real bar. Even where a standard is voluntary, it can become a precondition for doing business.

BPP works at the point these requirements create friction: Chinese-entity access without a Chinese licence, a maintained Chinese-language source of record, compliance-reviewed content built to produce the audit trail the current standards ask for, and tiered authoritative placement across Baidu Baike, Baijiahao and Baidu Zhidao.

The rules are arriving faster than the market expected. The brands that will be affected first are the ones that have been treating China's GEO standards as someone else's problem.