Attorney Shao Shiwei | Is GEO Classified as a Gray-Market Activity? An Analysis of the Compliance Boundaries and Legal Risks of Generative Engine Optimization

What Is GEO? The Evolution from SEO and ASO to Generative Engine Optimization

Within the evolutionary trajectory of digital marketing, Generative Engine Optimization (GEO) can be viewed as a natural extension of Search Engine Optimization (SEO) and App Store Optimization (ASO).

The core of SEO lies in studying search engine algorithms and employing tactics such as keyword placement, backlink building, and page authority enhancement to achieve higher rankings in search results. ASO emerged during the mobile internet era, whereby service providers optimize an application’s metadata—including its name, keywords, and description—to secure better positions in app store rankings. In the age of AI-powered search, although the underlying logic has shifted, the fundamental motivation remains consistent: merchants and brands still seek priority visibility for their information. The only change is that the “entry point” has shifted from webpages and app stores to AI assistants such as ChatGPT, DeepSeek, and Wenxin Yiyan.

The key distinction is that SEO delivers a “collection of links,” requiring users to click through and make their own selections, whereas GEO directly influences AI-generated answer recommendations. When users pose questions, they typically receive concise textual summaries. In other words, information delivery has shifted from “presenting multiple possibilities” to “providing limited answers.” Under these circumstances, those who succeed in becoming part of the AI’s “core corpus” gain more direct user attention. This explains why, despite the lack of transparency surrounding GEO rules, numerous teams and engineers transitioning from SEO and digital marketing have begun experimenting with this approach.

However, precisely because AI-generated answers carry a stronger sense of authority, boundary issues associated with GEO have rapidly come to the fore. Some actors choose to enter the AI’s “trusted source pool” by providing high-quality, verifiable content, while others attempt to “poison” AI systems by fabricating expert identities, inventing research reports, or covertly injecting prompts. This raises a critical question: Is GEO a new opportunity in digital marketing, or is it a risk-laden area prone to gray-market activities?

Is GEO Synonymous with Gray-Market Activities?

When discussing whether “GEO constitutes gray-market activity,” we must clarify a fundamental premise: technology itself is neutral. For merchants, GEO is a content and channel strategy oriented toward generative platforms, aimed at enhancing their brand’s visibility and the likelihood of being reliably cited in AI-generated responses. In other words, GEO possesses clear commercial and technical value at the conceptual level; the issue arises from “how service providers fulfill merchants’ demands.”

Compliant practices typically involve organizing and distributing a company’s genuine credentials, verifiable data, and structured information in accordance with AI crawling preferences. This amounts to rendering “facts” into machine-readable formats, representing a continuation of content optimization and brand digitization. By contrast, practices that create false consensus, forge authority, or covertly inject information to mislead model judgments are what the industry and regulators refer to as “gray-market” activities ordata poisoning. Therefore, the legality of a given GEO operation hinges on the methods employed and the subjective intent behind them, rather than on “GEO” as a concept.

The so-called "data poisoning" can be understood as the injection of distorted information into corpora that are visible to a model or may be retrieved by the model in the future, for the purpose of profit or competitive advantage, thereby altering the model's citation tendencies in its responses. Its operational logic is to generate a large volume of "seemingly natural" signals, causing retrieval and training mechanisms to misjudge the weight assigned to a particular brand or viewpoint. In the short term, such tactics may yield quantifiable exposure, but long-term risks primarily center on platform cleansing, reputational damage, and potential legal consequences.

In practice, common variants of "poisoning" tactics include:

• One approach relies on scale to generate "public opinion heat"—using automated tools to mass-produce Q&A-style content and experience posts, and leveraging a large number of accounts to disseminate such content into public corpora pools such as Zhihu, Xiaohongshu, and forums;

• Another variant operates through "domains": purchasing abandoned domains, publishing advertorial content, and then directing weight to target pages through redirects or by hosting parasitic articles on legitimate websites, thereby increasing the probability that such pages will be deemed "authoritative" during indexing;

• There are also technically more covert pathways, such as embedding guiding instructions or comments (gray-box prompt injection) in public documents or code platforms (e.g., GitHub, Notion), with the aim of having the model capture relevant information when it scrapes these public resources.

The common characteristics of these tactics are:

They attempt to leverage retrieval and statistical patterns to create an "illusion of consensus." However, it should be noted that mainstream AI platforms are continuously enhancing their detection capabilities for anomalous patterns — short-term abnormal traffic, collective behavior from accounts sharing the same source, and anomalous external link patterns may all trigger automated risk control measures, resulting in downranking of pushed content, account suspension, or entire-site blacklisting. Furthermore, once a platform labels a domain or brand as a low-credibility source, the cost of recovery often exceeds the short-term gains initially obtained.

In summary, GEO itself is not part of the gray market; however, if the technology is abused and GEO-related operations slide into gray areas, the resulting harm will not be limited to individual brands but will also affect users, the industry, and even the trust foundation of the entire AI platform. The following section will analyze the specific manifestations of these harms.

Harms Caused by Gray-Market GEO

On the surface, gray-market GEO can generate substantial exposure in a short period, delivering "immediate" traffic effects for clients. However, such traffic is often built upon false content and data manipulation, and the underlying harms extend beyond mere "content distortion," triggering chain reactions across multiple levels.

1. Harm to Ordinary Users
 

First, the most direct victims are the broad consumer base. When gray-market GEO introduces fabricated cases, exaggerated efficacy claims, or fictitious authoritative endorsements into AI responses, users often treat such information as objective advice. Particularly in fields such as medical aesthetics, financial wealth management, and health and wellness, this type of information directly influences consumer decisions, ranging from the purchase of inferior products to delayed treatment and financial losses. A deeper concern is that repeated exposure to similar content can easily trap users in an "information cocoon" derived from a single information source, gradually impairing their ability to compare and discern.

2. Impact on the AI Industry

Furthermore, such false content does not merely mislead users. Once massively incorporated into training corpora or retrieval pools, the accuracy of AI model responses becomes compromised, leading to biases and even logical inconsistencies. Over time, this erodes user trust in AI, potentially causing users to perceive AI as unreliable and thereby reducing their usage and reliance. This erosion of trust not only undermines the platform’s competitiveness but may also hinder the widespread adoption and practical implementation of the AI industry as a whole.

3. Damage to Merchants and the Industry Ecosystem
 

Finally, gray-market GEO practices harm legitimate merchants and the broader industry ecosystem. They often rely on “low-cost fabrication” to secure short-term traffic, thereby obscuring genuine, compliance-oriented content. Over time, this creates a “bad money drives out good” dynamic, marginalizing compliant enterprises while allowing fabricators to reap undue advantages. The ultimate consequence is not only the stigmatization of the GEO optimization sector as a “gray industry,” but also a drag on the reputation and public credibility of the entire digital marketing field. Moreover, once false advertising, data falsification, or traffic-diversion fraud is substantiated, the relevant enterprises and service providers may face civil liability and even criminal prosecution.

Potential Criminal Legal Risks for GEO Service Providers

Although GEO service providers do not directly provide funds, products, or transactions to their clients’ customers in the course of rendering services, if their optimization activities are combined with their clients’ illegal or criminal conduct, judicial authorities may characterize such activities as “providing technical assistance.” The following high-incidence offenses are those that GEO service providers and practitioners must be particularly vigilant about.

1. Crime of Aiding Information Network Criminal Activities (the “Aiding Crime”)

In judicial practice, any person who “knowingly provides support to others committing information network crimes” may incur liability for the Aiding Crime.

While GEO optimization appears to be an information service, if the provider knows that the client operates a Ponzi scheme, gambling website, or fraud platform, and nevertheless increases its exposure on AI platforms by enabling AI to crawl and index related content, such acts essentially facilitate traffic diversion to the illicit operation.

For example, in a certain Web3 Ponzi scheme project, the engaged GEO team crawled public corpora, performed vectorization, and embedded core tags such as “high returns, zero risk” into massive volumes of Q&A texts, thereby inducing AI systems to treat these claims as broad user consensus. This was used to induce retail investors to participate.

If the optimizing party continues to provide services despite knowing that the client’s fund operations are abnormal, it may be deemed to have committed the Aiding Crime. Although these acts appear outwardly as “optimization,” from a criminal law perspective they amount to providing “traffic diversion plus disguised endorsement” to a criminal organization, making them highly susceptible to characterization as the Aiding Crime.

2. Fraud

Where a generative engine optimization (GEO) service provider does not merely “provide optimization” but participates in conduct that helps to “fabricate facts and conceal the truth,” it may be deemed an accomplice to fraud.

For example, in the financial wealth-management sector, continuously feeding false information such as a “2025 High-Yield Wealth-Management Platform Ranking,” embedding actually unlicensed phishing websites into answers and packaging them as industry leaders, and promising “annualized returns of up to xx%,” after which users register and deposit funds and the platform absconds.

Once such content is disseminated through GEO, if users suffer property or personal harm due to reliance on AI recommendations, judicial authorities may determine that the optimization provider and its client constitute accomplices. The key point in such conduct is that the content itself involves fabrication or concealment, and the GEO service provider is deeply involved in the “counterfeiting chain,” with its criminal status closer to that of an “accomplice” rather than that of a “mere outsourced service provider.” By helping to adjust AI platform recommendation results, the provider is more easily drawn into the crime of fraud.

3. Crime of False Advertising

False advertising has long been prevalent in traditional marketing, and the amplification effect of GEO makes its risks more pronounced.

For instance, in the medical aesthetics industry, packaging the lead surgeon of an institution with false titles such as “Chief Expert of the Asian Anti-Aging Research Institute,” and using massive soft articles plus structured data feeds to cause AI, when answering “Who is the most authoritative provider of facial anti-aging treatments,” to prioritize citing that identity; alternatively, self-producing a “2025 Top 10 Comprehensive Strength Chinese Medical Aesthetics Brands” ranking, placing the client institution among the top three, and then centrally distributing it on portals, document libraries, and forums under the guise of a “white paper” or “industry report,” so that AI crawls and incorporates it into its answers.

Traditional false advertising is often confined to the platforms on which it is placed, whereas the problem with GEO is that it “directly enters AI’s answers,” resulting in broader audience coverage and stronger misleading effects. Under the Criminal Law, where advertisements are used to make false publicity about goods or services and the circumstances are serious, the responsible persons may face fixed-term imprisonment of up to two years.

4. Crime of Illegal Business Operations

Article 225 of the Criminal Law provides that those who, in violation of state regulations, engage in illegal business operations and disrupt market order, where the circumstances are serious, may be held criminally liable in accordance with the law. This article both enumerates several specific scenarios and includes a catch-all clause for “other illegal business operations that seriously disrupt market order,” thereby exhibiting expansiveness.

For example, paid post-deletion services have not infrequently been characterized as the crime of illegal business operations in judicial practice. For instance[i], in the Nanjing “Hacker Post-Deletion Case,” technical personnel illegally obtained website account credentials and provided post-deletion services to clients, and were convicted of the crime of illegal business operations by the Xuanwu District People’s Court of Nanjing; a certain marketing company in Beijing was found guilty of illegal business operations by the Chaoyang District People’s Court of Beijing for providing paid post-deletion services to clients.

GEO (generative engine optimization) services do not inherently possess a “paid post-deletion” function, but in the gray market, some service providers indeed package “making negative information disappear from AI answers” as a disguised form of post deletion, thereby crossing criminal red lines.

An interesting question is whether providing GEO “poisoning optimization” services can give rise to charges that the GEO service provider has committed the crime of illegal business operations.

Attorney Shao opines that if the core logic of GEO services is to suppress negative information and fabricate false consensus through large-scale "corpus poisoning," "targeted RAG feeding," and "prompt injection," such conduct may fall within the catch-all provision of the crime of illegal business operations ("other illegal business operations that seriously disrupt market order").

5. Crime of Opening a Casino

Gambling-related clients constitute a concealed yet frequent source of demand in the GEO industry. Whether for overseas gambling websites or online gambling platforms disguised with blockchain technology, if a GEO service provider offers optimization services that cause these entities to appear in AI-generated responses within contexts such as "recommended sports and entertainment platforms" or "popular online gaming platforms," this amounts in essence to providing traffic diversion and promotional services for gambling platforms.

In judicial practice, such conduct is often treated as complicity in the crime of opening a casino. The criminal risk is particularly high when GEO services use targeted feeding to guide a substantial number of domestic users to overseas gambling platforms. Common methods include using "domain farms" or "authority hijacking" to initially disguise gambling links by attaching them to legitimate entertainment websites, and then redirecting users to gambling pages, thereby deceiving the AI's "authoritative source" determination mechanism.

Pursuant to provisions such as the Several Opinions on Handling Cases of Cross-Border Gambling Crimes, entities that provide software development, technical support, advertising placement, member recruitment, and other services to gambling websites may be prosecuted as accomplices to the crime of opening a casino. In the GEO context, if optimization activities directly serve to attract gamblers and expand traffic, the service provider may be determined by judicial authorities to be an accomplice to the crime of opening a casino.

6. Crime of Infringing on Citizens' Personal Information

In the practical operation of GEO services, data serves as the underlying driver of optimization. To ensure that AI systems "prioritize citing" a particular brand when generating responses, service providers often need to construct user profiles, conduct vector analysis, and customize fed corpora. If this process relies solely on publicly available and compliant data, it remains within the scope of content optimization; however, in gray-market scenarios, many service providers cross boundaries by directly collecting or even purchasing privacy-involved data for model training and precise feeding.

Common practices include:

• Scraping private corpora via web crawlers: Using crawler programs to batch-collect medical consultation records and symptom descriptions from forums and medical Q&A communities, and processing such sensitive corpora into "real cases" fed to models to trigger recommendations for specific brands when users make inquiries;

• Trading black-market data packages: Purchasing loan application information, credit reporting data, and insurance customer lists to fabricate "successful financial management cases" or "posts on low-interest loan experiences," thereby increasing their weight in AI-generated responses;

• Collusive integration involving internal and external parties: Certain service providers collaborate with intermediaries or platforms to obtain real contact information, such as mobile phone numbers and WeChat IDs, through APIs or gray channels, and then embed such information into "recommended answers" to guide users to directly contact target merchants.

In these scenarios, user data is not merely "incidentally present" but serves as "raw material" used by GEO service providers to enhance optimization effectiveness. Once the volume of involved data reaches the threshold for criminal liability, or the data categories themselves constitute sensitive information (such as medical, financial, or location trajectory data), the conduct may constitute the crime of infringing on citizens' personal information.

A more insidious risk lies in the fact that such data rarely appears in the form of "direct buying and selling." Instead, service providers package it as part of "precision targeting" or "intelligent Q&A optimization," leading operators to underestimate the risks. However, from the perspective of judicial authorities, the mere existence of illegal acquisition and utilization satisfies the constituent elements for criminal liability.

In practice, when GEO service providers engage in areas such as Ponzi schemes, gambling, medical advertising, or the black market for data, they are highly likely to attract the attention of judicial authorities and may face criminal legal risks, including aiding information network criminal activities, fraud, false advertising, illegal business operations, operating casinos, and infringing upon citizens' personal information.

Furthermore, it is important to note that these charges are not exhaustive. In specific cases, provisions such as illegal intrusion into computer information systems, refusal to fulfill obligations for information network security management, and illegal use of information networks may also be applied by extension. In other words, the gray operational space of GEO often corresponds to more complex boundaries of legal risk, requiring practitioners to maintain a high degree of caution.

Challenges and Opportunities Coexist: The Way Forward for GEO Service Providers

It is certain that GEO will not disappear due to "controversies surrounding gray-market activities." As AI-powered Q&A gradually replaces search engines as the primary entry point, demand for related services will only grow stronger. Recently, U.S.-based Profound secured tens of millions of dollars in investment from top-tier venture capital firms such as Sequoia Capital, indirectly confirming that the GEO market is experiencing rapid expansion. This influx of capital indicates that the commercial value of this sector has been recognized, and the domestic market also holds significant potential.

At present, the domestic GEO landscape is still dominated by grassroots teams and SEO firms undergoing transformation, characterized by rudimentary operational models, insufficient market education, and a lack of clearly defined budgets among small and medium-sized enterprises. However, entities in high-margin industries such as finance, medical aesthetics, and education have begun to explore investments in this area. It is foreseeable that in the coming years, as regulations tighten and gray-market tactics are phased out, the service providers that endure will be those committed to authentic content and compliant operations. Those who pioneer compliant pathways will be better positioned to secure a competitive advantage as the industry matures.

[i] Cyberspace Administration of China Announces Ten Typical Cases of "Online Extortion and Paid Post Deletion" -- Current Affairs -- People's Daily Online http://politics.people.com.cn/n/2015/0126/c1001-26452824.html