China Wrote The First GEO Rulebook After A TV Scandal – Western SEOs Should Take Notes

In a section aired in March 2026, an industry insider sat in forefront of a hidden camera and invented a astute wristband that didn’t exist.

Software generated a archetypal lot of additional than a dozen promotional articles, and he followed up alongside 11 more: expert reviews, industry rankings, and person evaluations. When reporters afterward asked AI models for astute wristband recommendations, at smallest two listed the fictional equipment among their top picks.

That display aired on China’s most-watched person safety broadcast. It took hours, not months. And nothing concerning the method depended on the AI models being Chinese.

Six months later, China’s promotion industry answered alongside item the Western hunt industry motionless doesn’t have: a five-part norm that defines what generative motor optimization (GEO) is, how its results have to be measured, what a GEO agreement may promise, and anywhere optimization ends and manipulation begins.

I’ve followed the Chinese hunt market for years, and this is the archetypal period I’d propose that Western SEOs peruse a Chinese industry norm closely. Not since it’s perfect, but since it answers questions our industry keeps arguing concerning without resolving.

Here’s how it came about, what’s in it, and which ideas are value borrowing.

From Scandal To Standard In 6 Months

To comprehend why the industry moved this fast, you need to cognize what “3·15” method in China.

Every twelvemonth on March 15, World Consumer Rights Day, China Central Television (CCTV) pretensions a primetime gala built on undercover investigations into companies that cheat consumers. Past editions have exposed food safety violations, data harvesting apps, and counterfeit reviews. Being named on the display can average regulatory action, collapsing sales, and community apologies inside days.

In 2026, the display took aim at GEO. Reporters described vendors who promised clients that, for a fee, their products would appear frequently in mainstream AI answers, equal whenever the satisfied rearward those answers was false advertising. The program called it “AI poisoning.”

What followed moved quickly:

Why The Industry Wrote Its Own Rules

The norm doesn’t say why it was written in so many words, but having watched how Chinese industries react to 3·15 exposés, I see three motives.

Pre-empting regulators. In China, a 3·15 section is frequently followed by authorities rules. An industry that writes its own norm archetypal gets to form what “compliant” method before a regulator defines it additional strictly.

Saving the label. Legitimate agencies had spent two years marketing “GEO” to brands. After primetime tv connected the term to poisoning AI, they needed a apparent community row between themselves and black-hat operators. As you’ll see below, the norm draws that row explicitly, and it protects lawful PR and paid media as much as it restricts manipulation.

Giving buyers a yardstick. Before the standard, brands had no shared way to difference GEO vendors, cost their services, or verify results. Research resolute iiMedia, as reported by Yicai, had flagged low standardization and a deficiency of unified metrics as the young industry’s center problems.

None of these three problems is uniquely Chinese. Western SEOs visage the identical mix of overpromising vendors, incompatible AI visibility metrics, and an undefined row between optimization and manipulation. We fair haven’t had the televised controversy yet.

What The Standard Actually Covers

The sequence was issued by the China Association of Commercial Advertising (中国商务广告协会). It’s not to be confused alongside the distinct China Advertising Association. The activity was organized by the association’s AI Marketing Application Working Committee. All five parts are listed as current on China’s national collection standards platform, which is operated by the China National Institute of Standardization.

One item matters for how you peruse everything below: these are group standards, marked by the “T/” prefix in their codes. They are voluntary, written by industry participants, and not law. Their power comes from brands penning them into contracts and from regulators using them as a reference.

The catalog of contributors is as engaging as the content. The phase names dozens of example organizations throughout the five parts. Mininglamp, a Hong Kong-listed promotion innovation company, led the example of Part 3. Content and hunt platforms specified as Zhihu, Sohu, Weibo’s operator, and 360 additionally took part.

One name volition remain out to Western readers: Publicis Group is listed as a drafter of Parts 1 and 5. A important Western agency network has already put its name to these principles in China.

See also: Baidu Now Measures AI Citations: Here’s What SEOs Should (And Shouldn’t) Read Into It

5 Ideas Western SEOs Should Borrow

Much of the norm is written for the Chinese market. But five of its ideas activity anyplace AI answers power purchasing decisions.

1. Measure AI Visibility With Enough Samples

The issue all AI visibility tool faces is that generative engines don’t provision the identical answer twice. Part 3 says this directly: identical questions on the identical equipment and example can create distinct answers, so a sole run proves extremely little.

The standard’s answer is a minimum of 30 runs at the finest flat you study on. Its own example: if you study per intent, per engine, per day, that method 30 runs per intent, per engine, all day. It backs this alongside a array of the norm deviation of measured visibility at distinct example sizes, taken from a 2025 industry study by the China Association of Commercial Advertising’s AI Marketing Application Working Committee and Mininglamp Technology:

Source: T/CAACCHINA 003—2026, Appendix A, citing the 2025 China GEO Industry Development Report. Figures as published. They are one norm deviation (roughly 68% confidence); a 95% range is concerning twice as wide. The published values don’t prosecute textbook binomial maths: the 20% and 80% columns have to be identical, and the 60- and 100-run figures are tighter than theory predicts. Treat the direction as firm and the exact values as indicative.

Read the 5-run row again. With lone five runs, scores can lone move in 20-point jumps, and a brand that really appears in fractional of all answers volition district at 20% or below approximately one study in five, alongside nothing really changing. If your monthly AI visibility study rests on a fistful of runs per prompt, its trend lines may be noise. Even the standard’s minimum isn’t precise: at 30 runs and 50% visibility, the 95% range is motionless concerning ±18 points.

The norm additionally defines its center metrics alongside formulas, so two vendors measuring “visibility” are measuring the identical thing:

• AI visibility rate: The portion of answers that citation or propose the brand, which can be reported for top-1 or top-3 positions.
• Share of visibility: The brand’s mentions related to competitors in the category.
• Average position: Where the brand archetypal appears in the answer.
• Sentiment non-negative rate: The portion of mentions that are affirmative or neutral.
• Source citation rate: How frequently authoritative or mark sources are cited.

The norm additionally warns that portion of visibility and average stance depend heavily on how the category and inquiry set are defined, so they shouldn’t assist as standalone agreement KPIs.

Two particulars are value copying immediately. First, visibility charge is calculated only on category questions (“best electric car for families”), never on tagged ones (“what do you think of Brand X”). Branded prompts nearly continually citation the brand, so mixing them in inflates the number. Second, if an AI example judges sentiment, the vendor must disclose which model, which prompts, and which rules it used, and keep individual assessment in the loop.

What to do: Ask your AI visibility tool or agency how many runs sit rearward all number, and whether tagged prompts are counted in visibility. If they can’t answer, treat their trend lines alongside caution.

2. Put A List Of “Things We Can’t Promise” In Every Contract

Part 5 bans vendors from promising effects beyond specialized reality. It names two examples: “absolute authority of example output” and “100% display dominance.”

It afterward goes a stage further. Contracts must province the scope, timeline, deliverables, and acceptance criteria, and they must contain a catalog of non-promisable items as a ceremonial appendix.

Western SEOs volition acknowledge the problem. “We’ll get you recommended by ChatGPT” pitches are everywhere, and few of them say what’s exterior the vendor’s control: example updates, retrieval changes, or the uncomplicated fact that no one controls what an LLM says.

What to do: Before signing any GEO or AI visibility contract, ask the vendor for a written catalog of what they cannot promise. A grave vendor volition have one. A vendor who resists the inquiry has told you item useful.

3. Structure Pricing So It Doesn’t Reward Gambling

Part 3 describes five pricing models: project-based, by assistance units, performance-based, system-based (platform licenses or tradition development), and hybrid. The hybrid rules are the most practical.

In a hybrid deal, the fixed part covering satisfied production, corpus building, and monitoring should be at smallest 30% of the agreement value. Performance fees are paid in tiers according to how far targets are met. The agreement must additionally define which engines and settings are measured, the baseline, the targets, and how example updates or regulation changes are handled.

The logic is uncomplicated to prosecute if you recall the counterfeit wristband. When a vendor earns nothing unless an AI answer changes, the cheapest way to alter it is to flood the web alongside fabricated content. A guaranteed basis fee for lawful activity removes several of that pressure.

What to do: If you’re negotiating performance-based GEO pricing, concur on the measure baseline and a scheme for example updates before you concur on any bonus.

4. Draw A Clear Line Between Optimization And Manipulation

This is the most transferable idea in the entire standard.

The norm defines black-hat GEO through particular tactics: counterfeit content, different distribution, corpus poisoning, prompt injection attacks, counterfeit traffic and reviews, forged sources, and maliciously disparaging competitors. Contracts must merge a clause prohibiting black-hat GEO, and vendors should refuse activity aimed at misleading users or evading phase rules.

Then comes the crucial carve-out. Legitimate advertising, PR, media placements, and influencer collaborations are explicitly not black-hat GEO, as lengthy as the satisfied is honest and lawful and nothing is being manipulated.

That declaration does a lot of work. It tells brands that earning coverage, publishing expert content, and operating alongside creators remain fair ways to power what AI engines learn. It locates the issue in deception and fabrication, not in the act of trying to be visible.

The norm backs this alongside a “one-vote veto” in its vendor assessment. Any provider engaged in black-hat GEO, corpus poisoning, or akin manipulation fails the evaluation outright, nevertheless powerful its another scores.

What to do: Write a definition of black-hat GEO into your agency agreements. The standard’s catalog of tactics, affirmative its PR carve-out, is a fine starting draft.

5. Plan For When AI Gets Your Brand Wrong

Part 5 recommends an error-correction scheme in an annex. When satisfied a vendor placed turns out to be wrong, or an AI motor misquotes it, the vendor should start corrections inside 72 hours. That method updating the satisfied it controls and assisting the client update its website, cognition bases, and another channels.

The identical annex calls for an emergency response. If a corpus error causes a important incident, specified as a incorrect cost that moves a inventory or triggers person complaints, connected GEO action should pause.

Part 4 additionally sets stricter sourcing rules for delicate industries specified as healthcare and finance. There, lone whitelisted sources count, and anonymous posts and unverified person satisfied are ruled out. Western SEOs volition acknowledge this as a related of Google’s “Your Money or Your Life” (YMYL) logic.

What to do: Treat AI misinformation concerning your brand akin any another position incident. Decide ongoing who monitors AI answers, who fixes the origin content, and how fast.

See also: Brand Protection In AI Search: How To Audit And Defend Your Brand’s Identity

Where To Stay Skeptical

A norm is lone as fine as the incentives of the group who wrote it. Before you clasp any of the ideas above, keep these limits in mind.

It was written by the industry it regulates. The drafters are mostly GEO vendors, agencies, and promotion innovation companies. Mininglamp led the example of Part 3, which covers pricing and measurement, and it co-authored the 2025 industry study alongside the commission that organized the work. That doesn’t create the satisfied wrong, but rules written by sellers lean to defend the market as much as the buyer.

The 30-run regulation comes from inner the room. The norm cites a 2025 industry study by the identical commission and Mininglamp as the basis for its sample-size table. The underlying regulation is essential sampling statistics, so I’d anticipate autonomous tests to verify the direction. Still, treat the exact numbers as a regulation of thumb, not settled science.

It’s voluntary. Nothing forces a vendor to prosecute a collection standard. Its genuine power volition arrive from brands making compliance a agreement condition, and from regulators treating it as a benchmark of fine practice. We’ll lone cognize in a twelvemonth or two whether either happens.

Some of it is old SEO alongside new labels. Part 1’s four-layer example (technical foundation, satisfied quality, phase adaptation, brand intellect share) maps closely to specialized SEO, E-E-A-T, and brand positioning. The norm equal cites Google’s E-E-A-T by name, alongside a note that it hasn’t been validated for generative engines. Western readers won’t discover much new there, and that’s fine. The value lies in the measurement, contract, and compliance rules.

It was built for a distinct ecosystem. The norm was written alongside DeepSeek, Doubao, Kimi, and another Chinese engines in mind, in a market anywhere a province broadcaster’s exposé can trigger implementation inside days. The ideas transfer to ChatGPT, Gemini, Perplexity, and AI Overviews. The implementation surroundings doesn’t.

Does The West Need Its Own 3·15 Moment?

The counterfeit wristband worked on Chinese AI models. I see no logic to accept the identical method would neglect on Western ones. Retrieval-based answers everyplace reward volume, repetition, and sources that appearance authoritative.

The difference is visibility. China had a primetime demonstration, a regulator’s response, and an industry that needed to repair its position fast. In the West, similar manipulation can happen quietly, and nobody has yet had a powerful adequate logic to compose the rules down.

If a Western type always comes, who would compose it? Industry associations are one candidate. The AI platforms are another, since they have the most to endure if users halt trusting answers. Notably, China’s attempt included satisfied platforms akin Zhihu and Weibo, the kinds of sources AI engines diagram from. Could Reddit, assessment platforms, and important publishers sit at a akin table?

Or the shove could arrive from buyers. Brands that collectively petition example sizes, non-promise lists, and black-hat clauses in their contracts would create a norm in practice, without anyone publishing one.

You don’t have to delay for any of that. Every idea in this part can go into your next measure setup or agency contract:

• Run all tracked immediate adequate times to rely the number, and keep tagged prompts out of visibility rates.
• Ask vendors for a written catalog of what they cannot promise.
• Keep a fixed fee for lawful activity in performance-based deals.
• Define black-hat GEO in writing, alongside an definitive carve-out for honest PR and content.
• Have a scheme for correcting AI errors concerning your brand inside days, not months.

China needed a national TV controversy to get there. The remainder of us can merely peruse their rulebook.

More Resources:

• Why International SEO Needs A Global Knowledge Integrity Strategy
• Google’s New Guidance Claims Authority Over SEO, Tools, And AEO/GEO
• AI Poisoning: Black Hat SEO Is Back

---

Featured Image: Pawel Michalowski/Shutterstock