Poison AI to disrupt the advertising industry! Huang Shengmin, Director of the China Advertising Museum: Build a content bank and refuse to be kidnapped by GEO

Every reporter｜Song Xinyue Special reporter Wen Muxia Intern Chang Songzishen Every editor｜Gao Han

Recently, CCTV’s “3.15” gala exposed illegal businesses using GEO (Generative Engine Optimization) technology to feed false information into AI databases, igniting public debate on topics such as “poisoning AI.”

When seemingly objective AI responses are manipulated, the public can’t help but worry: are the products and services we rely on AI recommendations for just “traps” packaged in false information? This incident concerns not only technological safety but also affects various industries, with the advertising industry, closely tied to consumer decision-making, being the most severely impacted.

After all, when consumers ask AI “which face mask is worth buying,” no one wants to receive a carefully crafted false answer. AI was originally a new avenue for advertising and marketing but has been reduced to a tool for false promotion due to GEO abuse, misleading consumers and overextending the credibility of the AI information ecosystem, plunging the advertising industry into a trust crisis. How can we clarify the root of the chaos and find a way out? With these questions in mind, a reporter from “Everyday Economic News” (hereinafter referred to as “Everyday Reporter”) spoke with Huang Shengmin, founding dean of the School of Advertising at Communication University of China and curator of the China Advertising Museum, to understand the profound impact of GEO on the advertising industry and explore pathways for resolution.

Huang Shengmin Image source: Provided by the interviewee

Why can GEO poisoning hijack the advertising ecosystem?

When businesses use “black hat GEO” tactics to make AI recommend fictional products and turn false advertising into AI’s “standard answers,” users who trust AI complete their purchases only to be deceived—this behavior not only causes losses for consumers but also brutally disrupts the interactive closed loop between brands and users in advertising communication.

According to Huang Shengmin, this is the core harm brought to the advertising industry by GEO abuse. When such behavior scales, AI, once an information assistant, becomes a “toxic information room” filled with false content.

The rapid formation of an AI poisoning industry chain behind GEO is due to the irreversible migration of traffic entry points. Huang Shengmin provided the Everyday Reporter with a set of intuitive data: the total scale of China’s advertising industry will reach 15 trillion yuan by 2025, accounting for a reasonable range of 1% of GDP, with traditional media advertising only 50 billion yuan, while advertising related to digital media and AI will reach as high as 14.5 trillion yuan. As the daily active users of AI large models surpass 100 million, the habit of users obtaining information has shifted from search clicks to AI Q&A, leading companies to be anxious that “disappearing from AI’s standard answers means being voiceless in the next era,” which also provides soil for the rampant growth of GEO.

“The GEO being done now is actually a shift of the approach used in the past for SEO, where service providers use industrialized production of content garbage, diluting the training data for AI, betting on the randomness of sampling in large models,” Huang Shengmin stated.

Image source: DeepSeek

The core issue enabling GEO to achieve large-scale poisoning lies in the underlying logic of the “three infinities” constructed by digital technology and AI, along with the “computationalism” behind it. As early as 20 years ago, Huang Shengmin proposed the “three infinities” brought by digital technology: infinite content production, where anyone can write a hundred articles in a minute using AI; infinite transmission, where no channel can monopolize information due to satellites, 5G, Wi-Fi, etc.; and infinite reach, where people can access information anytime and anywhere through their phones.

These three infinities overturn the traditional advertising logic based on “optimal allocation of scarce resources,” becoming a breeding ground for GEO chaos. Meanwhile, computationalism—“everything can be computed; if there are things or events, there is data”—means that AI only recognizes the richness, relevance, and network weight of data but cannot distinguish between fabricated and real content.

Mass-produced soft articles and fictional ranking lists are just a series of grabable, computable 0s and 1s to AI. The continuous feeding of low-quality data traps AI in a vicious cycle of “low-quality data leading to low-quality output leading to even worse data,” continuously depleting the credibility of the entire AI information ecosystem.

Even more alarming is that GEO poisoning has thrown the advertising industry into a “bad money drives out good” dilemma. Some businesses, seeking short-term profits, use false information to attract attention, while those enterprises that focus on products and brands are overlooked due to the lack of such “traffic tricks.” Huang Shengmin candidly stated that the essence of advertising is the circulation of commercial information, and its core is to establish positive interactions between brands and users. “Any communication must consider whether the other party’s reception and feedback are positive,” whereas GEO poisoning completely violates this principle, causing advertising to lose its core meaning of establishing trust and putting the entire industry’s ecosystem at risk of desertification.

Building a “Content Bank” to Bring Traffic Operations Back to Cognitive Operations

In the face of the chaotic industry situation caused by GEO poisoning, Huang Shengmin told the Everyday Reporter that the underlying paradigm of advertising has not changed due to AI technology; only the tools and efficiency have changed. “Advertising has always been the circulation of commercial information. In ancient times, people shouted in the streets; later, it was written on paper and became artistic fonts. Now, through AI feeding, the form is changing, but the essence of the interactive closed loop remains unchanged.”

GEO itself, as an optimization tool, is not a monster; compliant “white hat GEO” can help brands sort structured information, allowing AI to understand brand value more accurately. The key to resolving the industry’s issues lies in defining the boundaries of technology, using rules and systems to constrain the abuse of technology, allowing the advertising industry to return to a healthy track.

#给AI投毒已成产业链#'s topic trending on social media Image source: Social platform

Building a “Content Bank” is the core solution Huang Shengmin proposed to address AI data pollution and regulate the content ecosystem. In his view, content is the core production factor in the digital age, just like coal and oil during the Industrial Revolution, needing rights confirmation, assessment, and trading to form a healthy industrial cycle, which is the core meaning of the “Content Bank.” Just as every piece of content produced by authors, whether news articles or product reviews, can be registered on technologies like blockchain for rights confirmation, when AI needs to fetch content, it can engage in compliant transactional pricing, allowing content to become real assets through repeated citation and valuation, rather than being casually misused and contaminated data garbage. This mechanism can curb the production of low-quality “garbage content” at the source, allowing AI’s database to return to its essence of quality and authenticity.

However, Huang Shengmin also admitted that the construction of a “Content Bank” is not merely a technical issue but a complex social engineering project that cannot be completed by a single large company nor supported by individual enterprises—otherwise, it is likely to form a new monopolistic tool rather than an open public infrastructure. It requires massive data existence and circulation, as well as the cooperation of large companies, industry organizations, and regulatory authorities, forming a societal consensus on regulations and technical standards to break down data barriers and make content rights confirmation and assessment systems become universal industry rules.

In addition to building a content ecosystem, the role transformation of the advertising industry is also urgent. Huang Shengmin pointed out that the future evolution direction of AI must be a stronger fact-checking capability, which means advertising companies can no longer merely act as “traffic intermediaries” but must transform into “cognitive operators” for brands. In the past, advertising companies only needed to help clients buy traffic and place ads, solving media issues, whereas in the future, advertising companies must either delve deeper into technology, possessing core data algorithm capabilities, using technology to create a real, high-quality information dissemination system for brands; or move towards scene spaces, truly understanding clients’ localized needs and international perspectives, making advertising communication align with users’ genuine needs, and rebuilding trust between brands and users.

At the same time, deep integration of industry, academia, and research is also a crucial support for breaking through industry challenges. Huang Shengmin candidly stated that there are structural deficiencies in universities’ follow-up on AI technology, and only by promoting the integration of industry, academia, and research, combining theoretical research from universities with technical practice from enterprises, can we cultivate professionals who adapt to the AI era for the advertising industry and inject core momentum into industry development.