315 Exposes AI “Poisoning”: A Business That Grew from Putian to Silicon Valley
TechFlow Selected TechFlow Selected
315 Exposes AI “Poisoning”: A Business That Grew from Putian to Silicon Valley
From Putian to Silicon Valley, SEO swapped one letter.
Navigating Web3 tides with focused insightsAuthor: David, TechFlow
Last night, China’s annual March 15 Consumer Rights Day (315) exposed a business built on GEO.
Its full name is Generative Engine Optimization. You can think of it as:
Paying AI to say good things about you.
How does it work?
Brands want AI to recommend them first when consumers ask questions. So they hire GEO service providers, who mass-publish promotional soft articles across the internet. When AI models crawl these contents, they treat them as factual information and recommend them to users.
A CCTV reporter used software called “Liqing GEO,” available for purchase on Taobao.
The reporter fabricated an intelligent fitness band, inventing absurd product features such as “quantum entanglement sensing” and “black-hole-level battery life.” The software automatically generated over a dozen promotional articles and published them online.
Two hours later, the reporter asked an AI model: “Could you recommend some smart health bands?”
The AI placed this entirely fictional band at the top of its recommendation list.
The company behind the software is Beijing Lisi Culture Media Co., Ltd.—a one-person operation with zero employees enrolled in social insurance for multiple consecutive years.
This tool—developed by such a company—fooled mainstream Chinese AI large language models within just two hours.
The 315 exposure revealed AI “poisoning,” but this business may be far larger than a single Taobao-listed software tool.
SEO and the “Putian Story”
First, this isn’t new at all.
In 2008, CCTV’s “News 30 Minutes” program ran two consecutive days of reports exposing Baidu’s paid search ranking system. Paying money could place your website at the very top of search results—even ahead of listings for counterfeit drugs.
At the time, this business was known as SEO—Search Engine Optimization.
The biggest buyers were private hospitals affiliated with Putian, Fujian Province. In 2013, Putian-affiliated hospitals spent RMB 12 billion on Baidu advertising—nearly half of Baidu’s total ad revenue that year.
Many unqualified medical institutions leveraged SEO to push themselves onto the first page of Baidu search results—displayed alongside top-tier hospitals—making it impossible for ordinary users to distinguish legitimacy from fraud.
It wasn’t until the 2016 Wei Zexi incident—where a university student clicked on a top-ranked Putian-affiliated hospital listing and died after treatment—that regulators legislated clearly: paid search results are advertisements.
Yet this did not eliminate the business—only formalized it. Putian-affiliated hospitals continued buying rankings, now with a small “Ad” label beside each result.
But adding a label doesn’t stop people from clicking.
The fundamental problem with search engines has never been whether results are labeled—it’s that users instinctively trust top-ranked results.
Today, users have migrated from search engines to AI, believing AI is more objective and immune to paid ranking manipulation. Yet whoever controls the information distribution gateway controls ranking sales.
The gateway changed; SEO merely swapped one letter to become GEO—the logic of selling rankings remains unchanged, word for word.
What changed is the price.
GEO: Loved by Capital Markets
A business that refuses to die is precisely what capital markets love most.
In September 2025, BlueFocus Communication Group—the largest marketing communications firm in China—invested millions of yuan in PureblueAI (Qinglan), a GEO startup.
Qinglan helps real brands optimize their rankings and recommendation rates in AI search results. Clients include Ant Group, Tencent Cloud, and Volvo.
The products are genuine, the company is legitimate, and its work focuses on helping AI better understand authentic brand information.
This stands in stark contrast to the Liqing AI poisoning exposed by 315. Liqing fabricates products, invents specifications, and feeds false information to AI; Qinglan uses real brand content to align with AI’s recommendation logic.
Yet from AI’s perspective, both follow the same technical path: publishing content across the internet for AI crawlers to ingest.
AI cannot tell which content is marketing and which is fabrication. This ambiguity is precisely what makes GEO so ethically murky.
When BlueFocus invested in Qinglan, GEO was still merely an industry term among marketers. Three months later, it became a stock market concept.
By late December 2025, BlueFocus’ stock hit its daily trading limit.
Securities firms rushed to hold conference calls interpreting GEO; research reports defined it as “the next-generation traffic gateway in the AI era.” Capital flooded in—not only buying BlueFocus shares, but also driving up stocks linked to digital marketing or AI themes. BlueFocus surged 132% over nine trading days; a wave of follower stocks doubled in value.
Source: Cailian Press
After the surge, these companies quickly issued risk disclosures:
“GEO-related business generates no revenue and has no material impact on operations.” BlueFocus itself acknowledged that AI-driven revenue accounts for only a tiny fraction of its total revenue.
The implication is clear: the stock price more than doubled—but GEO itself hasn’t yet generated meaningful profits.
By late January, BlueFocus’ share price rose from RMB 9.6 to RMB 23.3—a 143% gain in one month. At that moment, Chairman Zhao Wenquan announced plans to sell up to 20 million shares—worth approximately RMB 467 million at prevailing prices.
Public research reports estimate that China’s entire GEO industry generated around RMB 2.9 billion in revenue last year. BlueFocus’ market cap increase over one month alone far exceeded that figure.
The 315 exposure revealed a Liqing system poisoning AI—for a few hundred yuan. But the GEO concept’s run through China’s A-share market netted billions.
Whether it qualifies as “poison” remains debatable—but the money earned is real.
315 Calls It Poisoning; Silicon Valley Calls It Monetization
In January this year, OpenAI announced on its official blog that ChatGPT would begin displaying ads.
Free users and subscribers paying $8/month for ChatGPT Go will see ads; premium subscribers remain unaffected.
On February 9, ads officially launched. Some appear at the bottom of ChatGPT responses, marked with a small “Sponsored” label. Initial advertisers include Ford, Adobe, Target, and Best Buy.
Ask ChatGPT “What car should I buy?” and you’ll get a response—with Ford’s sponsored link appended below.
OpenAI stated explicitly: ads won’t affect ChatGPT’s response content. Responses are responses; ads are ads—they’re separate.
That sounds familiar, doesn’t it?
Baidu said the same thing years ago: paid rankings are paid rankings; organic search is organic search—they’re separate. Later, the top five search results were all ads.
OpenAI expects ads to double its consumer-side annual revenue to $17 billion. ChatGPT boasts over 800 million weekly active users—95% of whom are free users, i.e., the full ad audience.
Looking back at the supply chain exposed by 315: Liqing injects soft articles into AI systems to promote nonexistent products. OpenAI places sponsored content beneath AI responses to promote paid products.
One operates without platform authorization—called “poisoning.” The other signs contracts with platforms—called “monetization.”
What’s the difference to users?
One appears inside the response; the other appears beneath it. One carries no label; the other displays “Ad.”
315 targeted a few-hundred-yuan Liqing tool; A-share markets inflated the GEO concept into a multi-billion-yuan bubble; OpenAI plans to earn $17 billion annually from this model.
The same activity transforms—from “poisoning” to “monetization”—with pricing increasing tens of thousands of times.
In November 2023, researchers from the Indian Institute of Technology Delhi and Princeton University published a paper on arXiv titled “GEO: Generative Engine Optimization.”
This marked the academic community’s first formal definition of the concept.
From that paper’s publication to the 315 exposure: just over two years. In between came gray-market exploitation, venture funding, stock market speculation, insider share sales, and AI platforms themselves entering the ad business.
The path SEO took over two decades, GEO traversed in just two years.
The difference? Back then, it took years for users to learn not to fully trust search engine results. Today, AI remains in its “trust dividend” phase—most users haven’t yet realized AI answers can be bought.
Still, this dividend period may not last long. Next time you ask AI “What should I buy?”, pause for one extra second:
Answers may be free—but your judgment shouldn’t be outsourced.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
深潮TechFlow
