The GEO Gold Rush Is Real — So Is the Mountain of Bullshit

Here's a number that should make every marketer on the planet sit up straight: China's GEO market is projected to grow from 600 million yuan in 2025 to 8.9 billion yuan in 2026 — and hit 50 billion by 2030. Nearly 100x growth in five years.

If you're not familiar with GEO (Generative Engine Optimization), think of it as what happens when AI search eats traditional search's lunch. Instead of competing for position on a results page, you're competing to be the source an AI model cites when answering a user's question directly. No clicks. No landing pages. Just your brand, embedded in someone else's answer.

The opportunity is enormous. The grifters have also arrived. Let me try to separate the two.

The Data That Should Terrify Anyone Still Doing SEO

In 2024, roughly 70% of Google's top-10 ranked pages were being cited by AI systems. By 2026, that number has cratered to under 20%. Read that again: even if you're ranking on the first page of Google, the AI probably isn't reading your content.

This isn't a failure of SEO strategy. It's a fundamental rewiring of how information gets delivered. Gartner projects that 68% of Google searches in 2026 result in zero clicks — the user's question gets answered on the results page itself, by AI. Your website can rank #1 and still get zero traffic.

In China, the shift is even more dramatic. ByteDance's Doubao AI assistant has 180 million monthly active users. 90% of China's top smartphone manufacturers have switched their built-in AI search engine to Doubao. When someone asks their phone "which restaurant near me is good," the answer doesn't come from a search ranking — it comes from what the AI decided to generate.

The B2B conversion rate for AI search is 14.2% — five times higher than traditional search. That's because users arriving through AI answers have already been pre-qualified by the model's understanding of their intent. They're not browsing. They're ready to decide.

The Grift: How GEO Became the New SEO Snake Oil

Here's where it gets ugly.

An article making the rounds this month — titled, bluntly, "GEO Collapses as Costly Black Box" — documents a mass exodus of small businesses fleeing GEO service providers. The complaint is universal: they paid thousands (sometimes tens of thousands) of dollars for "AI visibility optimization" and have absolutely no idea if it worked.

This isn't just a Western problem. China's GEO services market is even more unregulated. There's no certification, no standardized measurement, no accountability. Anyone can call themselves a "GEO expert" and start charging monthly retainers.

The business model is simple: sell anxiety to people who don't understand AI, deliver reports they can't verify, and hope they don't ask too many questions. It's the exact same playbook that made "SEO agencies" infamous in the early 2010s — except this time the technology is more opaque, so the grift is even easier.

But here's the thing: the existence of frauds doesn't mean the underlying opportunity isn't real. P2P lending collapsed in a wave of fraud too. That didn't mean fintech was worthless. It meant the frauds were worthless.

What Real GEO Actually Looks Like

Strip away the scammers and what's left? Three things that actually matter.

Structured data is not optional — it's the price of entry.

Schema.org markup isn't a nice-to-have. It's how AI models understand what your content is about, what category it belongs to, and what your core data points are. Early adopters with proper structured data are seeing 3-5x higher AI citation rates. This isn't magic — it's engineering. AI models need machine-readable signals. If your content is just walls of text wrapped in nested divs, you're making the AI work too hard to parse you. It won't bother.

The cost of implementing Schema.org is low. The technical difficulty is manageable. Most GEO agencies won't tell you this because it's too basic to justify their monthly fees.

Credibility infrastructure is the moat.

Xinhua News Agency and Fudan University recently published a GEO industry standard in China, built around a concept called "Three-Zone Governance" — strict separation of facts, opinions, and advertising. This matters enormously.

AI models are getting better at assessing source credibility. Content stuffed with promotional language, unverifiable claims, or hidden advertising gets downranked or ignored entirely. Content that demonstrates factual accuracy, proper attribution, and clear expertise gets prioritized.

This isn't something you can outsource to a content mill. It requires a systematic approach to how your organization produces and publishes information. It's infrastructure, not a campaign.

Multimodal is the next frontier.

AI systems are beginning to cite video segments, not just text. If your brand only has written content, you're half-invisible in multimodal AI search. The companies that figure out how to make video, images, and audio citable by AI will dominate the next phase.

Google Just Told You Something Important

Google officially stated they don't need llms.txt files — those special text files some GEO consultants have been selling as "essential for AI discoverability."

Think about what that means. The world's largest search company is saying that the way AI accesses information is fundamentally different from how traditional crawlers work. If you're still thinking about AI optimization in terms of feeding text files to bots, you're solving the wrong problem.

The real game is about entity authority, semantic structure, and trust signals — not file formats.

The Opportunity Is Real. The Execution Is Hard.

Let me be direct about what I think is happening:

GEO is going through a classic hype cycle. The money is pouring in, most of it is going to the wrong places, and a backlash is building. The contrarian article about GEO "collapsing" isn't wrong about the symptoms — but it's wrong about the diagnosis. What's collapsing isn't GEO itself. What's collapsing is the black-box vendor model that charges enterprises for results they can't measure.

The real opportunity sits in infrastructure: tools that make structured data deployment trivial, systems that measure AI citation rates transparently, platforms that help organizations build verifiable credibility at scale. This is boring, unsexy work. It's also where the 50 billion yuan market will actually be created.

If you're a business leader reading this, here's my honest advice: stop looking for "GEO agencies." Look for teams or tools that can restructure your content for machine readability, build your entity authority, and give you transparent metrics on AI citation. If someone can't show you exactly what they changed and how to verify the result, walk away.

If you're a builder looking at this space: don't build another service agency. Build the tools. The market doesn't need more humans selling GEO. It needs infrastructure that makes GEO accessible to every business, not just the ones who can afford a 30,000-dollar monthly retainer.

The Bottom Line

The traffic is moving. From search bars to chat windows. From blue links to generated answers. From clicks to citations. This shift isn't theoretical — it's happening right now, at scale, across every major market.

The companies that figure out how to exist inside AI's answer layer — not as a link to click, but as a source to trust — will own the next decade of digital visibility.

The grifters will burn out. The hype will cool. But the underlying shift is permanent.

Build for that.