Photo: VCG

After the China Central Television (CCTV) annual 3.15 Consumer Rights Gala spotlighted the so-called "AI poisoning" driven by a commercial technique known as generative engine optimization (GEO), some start-up companies in the GEO industry issued statements on Monday, vowing public compliance in AI search optimization.

CCTV finance channel, citing industry players, reported on Sunday that some GEO service providers promised clients that, for a fee, their products can frequently appear in mainstream AI-generated answers - even allowing false advertising content to be presented as "standard answers" in AI responses.

The Gala disclosed activities such as false information generation, AI data poisoning, ranking manipulation and malicious competition using GEO, which is seen as a growing threat to the integrity of large AI models, Jiemian News reported.

On Monday, Henan Henghui Hehuan issued a statement on its official WeChat account, saying its "Zhaixing GEO" product adheres to the principle of "technology for good, compliance first" and that the company "resolutely draws a clear line between itself and illegal black‑ and gray‑market operators."

According to the statement, the company pledged to strictly control content inflows, firmly reject the use of digital marketing techniques or fake traffic to boost rankings, and "will not cooperate with, participate in or tolerate any non-compliant GEO behavior." The statement also called on industry peers to uphold integrity, operate compliantly, compete fairly and jointly purify the AI and search ecosystem.

A Shanghai-based company, ABKE, also released an emergency statement, declaring its opposition to any form of false promotion, data falsification or malicious manipulation of AI outputs. ABKE's statement said the company does not participate in, develop or provide services described in industry reports as "brainwashing AI," "manipulating standard answers" or "one‑week ranking boost" black‑market offerings.

The CCTV report revealed how GEO tools, originally developed to optimize information release and improve promotional efficiency, can be repurposed. It quoted some industry players as saying that if large volumes of systematically targeted false information are placed online, they can be picked up by AI training and retrieval systems and subsequently surface as high‑priority answers to user queries. 

Experts define "AI poisoning" as the deliberate feeding of fabricated information - for example, fake expert identities or bogus research reports - into the information ecosystem so that AI systems learn and reproduce those falsehoods. Such actors have been labeled "poisoners" for their potential to contaminate corpus data.

GEO services are particularly insidious because it can disguise commercial promotion as apparently objective AI‑generated knowledge, making it hard for consumers to detect commercial intent, Pan Helin, a Beijing-based veteran economist, told the Global Times on Monday.

"Such behavior may infringe on consumers' right to informed choice and could violate advertising regulations on ad identifiability; fabricating facts may also amount to false advertising or unfair competition. If left unchecked, these practices risk distorting market competition and undermining public trust in AI as a reliable information source," Pan noted.

The training data of Chinese large models mainly come from the Chinese internet. Solving this problem requires joint efforts from multiple parties. If the issue is not taken seriously, speculators using GEO technology may rapidly pollute large models, Pan said.

According to Pan, authorities should actively improve the internet information ecological environment, enhancing internet information quality and increasing the supply of objective, neutral and scientific information.

In January, the State Administration for Market Regulation explicitly identified AI-generated advertising as a key challenge for internet advertising regulation in 2026, urging strengthened supervision and management of such practices.