China’s First GEO Trust Standard Just Took Effect: What Changes

China published its first trusted-GEO standard on August 11, 2026, and it took effect the next day. The document is called Generative Engine Optimization (GEO): Trusted Information Dissemination and Information Ecosystem Governance Specification (T/CAPT 026-2026), and it was drafted by Xinhua’s own Future Media Research Institute together with more than 30 institutions, including Tsinghua, Renmin University, Fudan, Beijing University of Posts and Telecommunications, Shanghai Jiao Tong, and the Chinese University of Hong Kong’s Shenzhen campus. JD’s own ad growth team sat on the review panel. You can read the original announcement on Xinhua’s site.

It is not a law. It is a group standard (团体标准), the kind an industry body publishes before a regulator turns it into something mandatory. But read who is behind it and that distinction stops being reassuring.

Why this is different from the usual GEO advice

Most of what gets written about GEO, ours included, is tactical: structure your content, get cited, build authority signals. This standard is the first attempt to write down what “trustworthy” actually means at the infrastructure level, and it does that with one idea that changes how a brand should organize its own content library.

The standard calls it “three-zone separation.” A brand’s knowledge base has to be split into three distinct stores: a fact library, an opinion library, and a marketing expression library. Any factual claim used in marketing content has to trace back to the fact library version of that claim. You cannot phrase your way around a fact in the marketing copy. If your fact library says a product passed a specific certification on a specific date, your ad copy cannot round that up, soften it, or dress it in a superlative that isn’t in the source record.

The standard also gives the industry its first formal line between white-hat and black-hat GEO. Content poisoning, fabricated consensus, answer manipulation, and prompt injection attacks against AI models are now named and banned, not just frowned upon. Sources get graded A through D for credibility, and every claim gets a traceability chain back to where it came from.

We have seen this movie before. Baidu’s search quality guidelines started as informal webmaster advice, then became the enforcement basis for entire ranking penalties once black-hat SEO got expensive enough for Baidu to care. A group standard with Xinhua, three top-tier universities, and JD’s own team behind it, explicitly designed with room to graduate into a national standard, reads like the early draft of that same playbook for AI search.

What is still unresolved: enforcement. The standard sets response-time targets (originally a four-hour containment window for high-risk incidents, and reviewers were already pushing to compress that further) and a three-tier capability system for GEO service providers, from L1 basic to L3 comprehensive governance. None of that means anything yet without a body actually checking compliance. That part has not been announced.

Who this actually affects

If your brand or your GEO provider is doing anything from the black-hat list, this is your notice, not a distant policy update. If you are simply not doing GEO yet, the standard is less urgent today and more of a preview of the bar you will be measured against once it does become mandatory or once platforms start referencing it in their own trust scoring.

The brands with the most to gain are the ones who get their fact library built before “trust-driven” becomes the industry’s default sales pitch. We worked with a mid-size skincare importer earlier this year whose entire product page consisted of adjectives: “premium,” “clinically proven,” “trusted by dermatologists.” Not one of those phrases pointed to an actual certificate, study, or date. When we rebuilt the page around a real, sourced fact block (which lab ran the test, which regulator issued the registration, when), DeepSeek started citing the brand by name in comparison answers within about six weeks. That was months before this standard existed. The standard just gives that approach a name and a rulebook.

The brands with the most to lose are the ones whose GEO strategy currently depends on volume: dozens of near-identical seeded posts, review farms, or an agency that promises guaranteed AI citations without explaining how. Under a source-tiering system, that kind of content gets graded down, not up, and it becomes the first thing a future compliance sweep would flag.

What to actually do about it

You do not need to wait for a national mandate to build the structure this standard describes. The mechanism is the useful part, independent of whether it ever becomes law.

Start with the fact library, because everything else in the standard hangs off it. List every factual claim your brand currently makes in China, in marketing copy, on packaging, in your Xiaohongshu notes, everywhere. For each one, attach a source: a certificate number, a lab report, a customs registration, a date. If a claim has no source, it is not a fact, it is marketing language, and under this standard it needs to be relabeled as such rather than presented as fact.

Second, separate that fact library physically from your marketing content, at least in how it is structured and published. Baidu’s Baijiahao, your own site, and any AI-facing content should point back to the same sourced version of a claim, not three different phrasings of it.

Third, audit whatever GEO tactics your team or your agency is currently running against the standard’s black-hat list. If nobody can explain in plain language how a tactic gets your brand cited, that is worth stopping before an auditor does it for you.

We cover the citation side of this in more detail, which AI platforms matter and how DeepSeek in particular decides what to cite, in our practical guide to appearing in DeepSeek AI answers. This standard is the compliance layer underneath that same work: build the fact library first, and the citation strategy in that guide has something real to point to. If you want a second pair of eyes on your current setup, our Baidu SEO and GEO team audits fact libraries and AI-citation readiness for foreign brands as part of the same engagement.

One thing worth arguing about: some brands will read a voluntary group standard and decide it is safe to ignore until it becomes law. Others will treat the institutional weight behind it, Xinhua, three top universities, JD, as a strong enough signal to act now, before the fact library becomes a scramble instead of a head start. Which side are you on, and what would change your mind?