Regulators and experts in China have warned of "artificial intelligence data poisoning" after the annual 315 consumer rights gala, which was held on Sunday, exposed the fabrication of promotional content to influence AI-generated response.

The findings of an investigation conducted by China Media Group, which were revealed during the live broadcast of the gala, showed the gross abuse of an online marketing tool called generative engine optimization, or GEO, that aims to enhance content visibility.

The probe revealed that companies have been using GEO to mass-feed promotional articles in an attempt to manipulate the information AI retrieves when answering user queries.

China Media Group reporters created a fake product — a smart wristband named "Apollo-9" — and claimed it had features such as quantum-entanglement sensors, a noninvasive blood glucose monitor and an ultra-long battery life.

Within hours of more than a dozen promotional articles about the nonexistent wristband being uploaded to a platform called Liqing GEO, two mainstream AI models began recommending the product in response to queries about smart wearable devices.

Li Fumin, a researcher of intelligent social governance at Shandong University of Finance and Economics in Jinan, the capital of Shandong province, said businesses that use GEO services to deliberately train AI models to promote specific products are essentially fabricating facts and engaging in bogus marketing.

"On the one hand, the practice leverages AI and algorithms to make false advertising, which results in unfair competition. On the other hand, this kind of behavior allows people to receive implanted marketing content without knowing it, which violates their consumer rights," Li said.

Generative engine optimization is widely considered the AI-era evolution of search engine optimization, or SEO. China's search engine companies, such as Baidu Inc, have previously faced scrutiny over paid ranking practices that blur the line between advertising and organic search results.

Several Chinese AI developers have responded cautiously to the investigation. ByteDance said that its chatbot, Doubao, was unaffected by such systems, while Alibaba said the core reasoning capability of its Qwen model was not affected by similar technologies.

Industry experts said the issue reflects a broader structural challenge rather than a flaw in any single model. They said that most AI systems rely heavily on accessible internet content, which can be mass-produced or manipulated.

Expediting legislation

Song Xiangqing, vice-president of the Commerce Economy Association of China, said that regulators should accelerate the formulation of new legislation to make deliberate pollution of AI data sources illegal.

China currently has no specific regulations governing GEO services. According to a work priority guideline on advertising recently released by the State Administration for Market Regulation, the nation will step up efforts to streamline digital advertising market strategies, including AI-generated content.

Song suggested drawing up a "white list" to identify trusted information sources and building a coordinated governance framework involving government oversight, corporate self-regulation and public supervision. "Without these safeguards, GEO services could evolve into a widespread source of information pollution, enabling data poisoning to spread throughout the AI ecosystem," he said.