315 Exposure: AI Data Poisoning, a Business That Runs from Putian to Silicon Valley

Author: David, Deep Tide TechFlow

Last night, the 315 Consumer Rights Day exposed a GEO-based business.

The full name is Generative Engine Optimization, which you can think of as:

Pay money to make AI speak favorably for you.

How does it work?

Brands want consumers to ask AI about them, and have AI prioritize recommending their products. So they find GEO service providers, who then mass-distribute promotional soft articles online. When AI fetches this content, it treats it as real information and recommends it to users.

A set called “LiQing GEO” software, available for purchase on Taobao, was used by CCTV reporters.

The reporter fabricated a smart wristband, inventing absurd selling points like “Quantum Entanglement Sensing” and “Black Hole-Level Battery Life.” The software automatically generated over a dozen promotional articles and posted them online.

Two hours later, the reporter asked AI: “Please recommend a smart health wristband.”

AI ranked this non-existent wristband at the top of the recommendation list.

The company behind this software is Beijing Lisi Culture Media, a one-person operation with zero insured employees for consecutive years.

This tool, created by a single company, fooled major domestic AI models in just two hours.

The 315 exposure revealed AI poisoning, but this business might be much larger than a Taobao software.

SEO, the past of Putian

First, this is nothing new.

In 2008, CCTV’s “News 30 Minutes” exposed Baidu’s paid ranking for two consecutive days. Paying money could get your website to rank first in search results, even for fake medicines.

At that time, this business was called SEO—Search Engine Optimization.

The biggest buyers were Putian private hospitals. In 2013, Putian hospitals spent nearly 12 billion yuan on Baidu ads that year, nearly half of Baidu’s total ad revenue.

Many unqualified medical institutions used SEO to push themselves to the first page of Baidu search, making them look like top-tier hospitals, indistinguishable to ordinary people.

It wasn’t until the 2016 Wei Zexi incident, when a college student who clicked on a top-ranked Putian hospital and died, that regulators legislated: paid search is advertising.

But this didn’t eliminate the business. It only formalized it—changing from gray-area activity to legitimate business. Putian hospitals still buy rankings, but now with a small “Ad” label next to the results.

However, once labeled, people still click.

The fundamental problem with search engines has never been whether there is a label, but that users naturally trust the top results.

Now, people have shifted from search engines to AI, believing AI is more objective and less polluted by paid rankings. But whoever controls the information distribution portal can sell rankings.

The portal has changed, and so has SEO—now called GEO, with the same logic of selling rankings.

What has changed is the price.

GEO, a darling of the capital markets

An unstoppable business, favored by capital markets.

In September 2025, China’s largest marketing firm BlueFocus invested millions of yuan in a GEO company called PureblueAI.

Pureblue helps real brands optimize their ranking and recommendation rates in AI search results, with clients including Ant Group, Tencent Cloud, and Volvo.

The product is real, the company is real, and they aim to make AI better understand brand information.

This is completely different from the AI poisoning exposed by 315. LiQing fabricates products, distorts parameters, and uses false info to deceive AI; Pureblue uses real brand content to align with AI’s recommendation logic.

From AI’s perspective, both paths are the same: publishing content online for AI to fetch.

AI can’t distinguish marketing from deception. This is the most ambiguous part of the GEO business.

When BlueFocus invested in Pureblue, GEO was just an industry term in marketing circles. Three months later, it became a stock concept.

By late December 2025, BlueFocus hit the daily limit-up.

Brokerages began holding conference calls to interpret GEO, calling it “the next-generation traffic portal in the AI era.” Capital flooded in, not only buying BlueFocus but also all companies related to digital marketing and AI concepts. BlueFocus’s stock surged 132% in nine trading days, and many concept stocks doubled.

Image source: Cailian Press

After the surge, these companies issued risk warnings:

GEO business has no revenue and does not significantly impact operations. BlueFocus also admitted that AI-driven revenue accounts for a very small part of total income.

In other words, the stock price doubled, but the GEO business itself hasn’t made much money.

At the end of January, BlueFocus’s stock price rose from 9.6 yuan to 23.3 yuan, a 143% increase in a month. At this point, Chairman Zhao Wengquan announced he would reduce holdings by no more than 20 million shares. Based on the current price, this would cash out about 4.67 billion yuan.

Public research reports show that last year, the domestic GEO industry market size was about 2.9 billion yuan. The market cap increase of BlueFocus in just one month far exceeded this figure.

315 exposed LiQing’s system poisoning AI with a few hundred yuan. But the GEO concept in A-shares made billions.

Whether it was poisoning or not is uncertain, but the profits are real.

315 called it poisoning; Silicon Valley calls it commercialization.

In January this year, OpenAI announced on its official blog that ChatGPT would start selling ads.

Free users and the $8/month “Go” users will see ads, while paid subscribers are 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, Best Buy…

If you ask ChatGPT which car to buy, it will give an answer with a Ford sponsorship link attached.

OpenAI clearly states: ads won’t influence ChatGPT’s responses. Responses are responses; ads are ads, separate.

Does that sound familiar?

Baidu also said the same back then. Paid rankings are paid rankings; organic search is organic. They were separated. Later, the top five search results were all ads.

OpenAI expects ads will double its consumer-side annual revenue to $17 billion. With over 800 million weekly active users, 95% are free users, all potential ad audiences.

Looking back at the 315 exposure of LiQing poisoning AI, it cost only a few hundred yuan. But the GEO concept in A-shares made billions.

Whether it was poisoning or not is uncertain, but the profits are real.

315 called it poisoning; Silicon Valley calls it commercialization.

In January 2024, OpenAI announced that ChatGPT would start selling ads.

Free users and $8/month “Go” users will see ads; paid users won’t.

On February 9, ads went live. Some appear at the bottom of responses, marked “Sponsored.” Major brands like Ford, Adobe, Target, Best Buy are involved.

OpenAI says ads won’t affect responses. Responses are responses; ads are separate.

Does that sound familiar?

Back in 2008, Baidu also claimed the same. Paid rankings are paid; organic results are organic. Later, the top five results were all ads.

OpenAI expects ads to double its revenue to $17 billion annually. With over 800 million weekly users, 95% are free, all exposed to ads.

Looking back at the 315 exposure: LiQing’s soft articles cost only a few hundred yuan, but the GEO concept in A-shares earned billions.

Whether it was poisoning or not, the profits are real.

The same event, from poisoning to commercialization, increased in price by tens of thousands of times.

In November 2023, researchers from IIT Delhi and Princeton published a paper titled “GEO: Generative Engine Optimization” on arXiv.

This was the first formal academic definition of the concept.

From the paper’s publication to the 315 exposure, just over two years. During this time, gray industry, financing, concept stock surges, chairman cash-outs, and AI platforms directly selling ads all happened…

The path SEO took twenty years ago, GEO completed in two years.

The difference is that back then, it took years for people to learn not to fully trust search engine results; now, AI is still in its trust dividend period, and most people haven’t realized that AI responses can also be bought.

But this dividend may not last long. Next time you ask AI what’s worth buying, remember to think for a second:

Answers can be free, but your brain can’t be outsourced.