Introduction

Recently, at the invitation of Longyun Co., Ltd., I gave a legal sharing on GEO (Generative Engine Optimization). After chatting with several big names in the industry, I got some new inspirations, and I would like to share them with you.

In the past two decades, the traffic distribution logic of the Chinese Internet has always revolved around the core action of "search". From the early days of "Baidu Click" to the later Site searches on WeChat, Xiaohongshu and other platforms, they were all extensions of the "Baidu Click" behavior, and gave rise to a mature SEO (search engine optimization) industry.

Now, the trend is quietly changing. Users are increasingly accustomed to directly asking questions to AI: "For a 30-year-old woman to fight against aging, should she choose an ultrasound cannon or a Thermage?" or "Recommend a bar suitable for watching football games."

The traffic entrance is moving from the "search box" to the "dialog box". When generative AI can bypass massive links and directly generate final answers for users, if it is not mentioned in the answer, it means that it is lagging behind the new era to some extent. This is why GEO has become the focus.

As legal practitioners, while we pay attention to the business opportunities, we also need to be clearly aware of the legal risks behind them. The evolution of technology often precedes the establishment of rules, and the GEO field has presented many gray areas that require careful legal definition!

Who’s in? Three major groups seize the new GEO continent

Although this is a brand-new field, it contains unlimited room for imagination - in the current highly involved market environment, new traffic entrances often mean lower customer acquisition costs and better competition opportunities.

As a lawyer who has long been concerned about the fields of Web3 and AI, I have observed that there are at least three major groups actively participating in them:

1. User: Provider of physical goods and services

They focus on the direct commercial conversion brought about by AI traffic, trying to gain priority exposure opportunities by influencing AI recommendation results.

For example:

• Medical beauty institutions abandon traditional search bidding and instead purchase "AI semantic injection tools", striving to give priority to AI in recommending their own institutions when users ask about "the best rhinoplasty doctor."

• Industries such as training institutions and automobile sales are also trying to use generative engine optimization (GEO) to let AI take the lead in recommending their products or services when answering relevant questions.

2. Investors: investment institutions and funds

They lay it out from two levels:

• Discover the track: By observing which companies have advantages in AI recommendations, we can determine their industry competitiveness and thus identify potential investment targets.

• Competing for the right to speak: Whoever can influence the corpus and recommendation logic of AI will be able to take the initiative in future investment advisory recommendations and industry analysis.

3. Service providers: GEO industry practitioners and entrepreneurs

This group of people usually have the ability to learn quickly and apply technology, and are actively involved in tool development, policy services and traffic operations. They continue to explore the boundaries and possibilities of this industry in various forms - some are positive innovations, and some are wandering in gray areas. This is the group that the second part of this article will focus on.

GEO’s three postures: huge profits, traps and legal red lines

In GEO's practice path, different methods are usually classified into three colors: "black, gray and white". As a lawyer, I must emphasize one point: the logical end of technology is often the starting point of law.

1. Black Hat: "Technical Manipulator" walking in the minefield

Typical means dismantling:

• Indirect Prompt Injection: Embedding instructions (such as white text) that are only identifiable by AI and invisible to human eyes in web pages, inducing AI to give priority to recommending specific content when answering.

• Knowledge Poisoning (RAG/Knowledge Poisoning): By injecting false or biased data, the public index database is polluted, causing AI to output preset biased results during the retrieval augmentation generation (RAG) process.

• False entity forgery: Forging address, qualifications and other information in public data sources such as maps and encyclopedias, polluting AI training data or real-time retrieval content, and creating a false reputation.

• Negative GEO attack: Implanting malicious code or sensitive words into competitor websites, triggering the AI security filtering mechanism, causing them to be blocked or marked as untrustworthy sources.

Legal risk characterization:

• Criminal level: It is extremely easy to constitute the crime of "destroying computer information systems" (Article 286 of the "Criminal Law"). Once the normal operation of the AI system is interfered with, the criminal red line is reached.

• Civil level: It is an obvious act of unfair competition (Article 11 of the Anti-Unfair Competition Law), and you need to bear liability for damages. The amount of compensation may be significantly amplified due to the spread of AI.

2. Gray Hat: the “traffic porter” walking on the edge

Gray hats try to avoid obvious illegal crimes, rely on scale effects to influence AI judgment, and believe in "quantitative changes lead to qualitative changes."

Typical means dismantling:

• Batch cleaning and semantic dimensionality reduction: Use AI to generate massive amounts of low-quality repetitive content, diluting real information and forcing AI to capture preset positive corpus.

• Simulated click stream attack (Bot-driven Interaction): simulate user click behavior through automated scripts, artificially increase the click-through rate (CTR) of specific content in AI, and defraud the algorithm weight.

• Masked Promotion: Organizes vest accounts to publish promotional content disguised as real experiences in batches on social platforms, so that it will be regarded as "user feedback" by AI and included in the search database.

Legal risk characterization:

• Liability for false publicity: This type of behavior essentially constitutes false publicity and violates the Advertising Law and the Anti-Unfair Competition Law. Regulatory authorities have gradually adopted the principle of "substance over form" to crack down.

• The risk of a brand being "blacklisted": Once identified by the anti-cheating system of the AI platform, the relevant domain name or brand may be permanently listed as an untrusted source, leading to its "digital death" in the AI environment.

3. White Hat: long-term value builder

The core of the white hat strategy is not to "manipulate AI", but to "become a high-quality data source trusted by AI". Although compliance costs are high, their accumulation has a significant compounding effect.

Typical methods include:

• Content structuring and summary optimization facilitate AI understanding and extraction;

• Deploy structured data (Schema Markup) to enhance the semantic clarity of content;

• Strengthen citations & factuality to improve information credibility;

• Adopt FAQ modeling to directly respond to common user questions.

We strongly recommend this path - it is based on compliance and wins the long-term trust of AI and users by continuing to provide authentic, high-quality, and verifiable content.

Looking at GEO from SEO jurisprudence: history does not repeat, but the illegal logic is the same

Although there is currently no judicial case specifically targeting GEO, it essentially has many things in common with SEO. Relevant judgments in the SEO field in the past are likely to become important references for future GEO cases. Below we combine several typical cases for analysis:

Case 1: Interfering with the Algorithm “Wanci Dominate the Screen” Case

In the SEO era, "Ten Thousand Words Dominate the Screen" was once a typical black hat technique: generating a large number of spam pages through high-authority websites and forcibly occupying keyword search results. In a related case, the court determined that this behavior disrupted the normal order of the search engine and constituted unfair competition, and ordered the defendant to compensate Baidu 2.753 million yuan.

Inspiration for GEO:

Some of the current GEO methods are the same, such as using AI to batch generate low-quality content and trying to "feed" the model to achieve answers that dominate the screen. This kind of behavior may not only cause the brand to be blacklisted by the model, but may also be legally recognized as "interfering with the normal operation of network products" and constituting unfair competition.

Case 2: Purchasing competitor keywords

In the “Fischer” trademark case, the defendant set someone else’s registered trademark as a search keyword, so that users’ search results would point to its own products. The court determined that this behavior violated the principle of good faith and constituted unfair competition.

Inspiration for GEO:

Similar logic may appear in GEO as more covert "prompt word injection" - for example, embedding inductive instructions for competing products in web pages in an attempt to influence the AI's answer orientation. This kind of behavior that indirectly misleads users and hijacks traffic through technical means may also touch the red line of anti-unfair competition.

Case 3: False question-and-answer word-of-mouth marketing case

Companies have previously been punished for organizing false “user experience” content on platforms such as Zhihu and Tieba. Regulatory authorities determined that such behavior deceived consumers, disrupted market order, and violated the Anti-Unfair Competition Law.

Inspiration for GEO:

Nowadays, some gray hat GEO methods are highly similar to this: Using AI to batch generate fake reviews and fake grassroots, to create a false reputation of "recommended by the entire network". It is necessary to clearly realize that AI is just a tool. If its output content is based on false information, it is still false propaganda, especially in fields with strong supervision such as medical beauty and health. The risk is extremely high.

Industry Compliance Warning: Different tracks, different “minefields”

To carry out GEO practice, we must combine the characteristics of industry supervision and see through the technical surface to see the bottom line of compliance. For example:

• Education and training: It is strictly prohibited to use corpus injection and other methods to use AI to make results promises such as "guaranteeing a pass" and "promoting the first score". As long as the content comes from its own feeding, the organization is the responsible party.

• Medical beauty institutions: belong to the category of medical advertising and require strict review. If GEO is used to induce AI to output efficacy comparisons, real-life cases, or disguised recommendations, it may directly violate medical advertising regulations. At the same time, we should also prevent competitors from using "negative GEO" to conduct business defamation.

• Big Health and Web3: Claims of curative effects and promises of high returns are all sensitive red lines. If the GEO strategy causes AI to output "zero-risk, high-yield" content, it can easily be suspected of false propaganda or even illegal operations.

The rise of GEO: mankind’s renewed struggle for information distribution rights

Based on industry observations, share the following views and suggestions:

1. Inspiration for entrepreneurial teams: Rather than waiting, it is better to go first

Although major Internet companies have resource and data advantages, their internal bureaucracy and standardized processes often respond slowly in agile and refined operation scenarios such as GEO. Therefore, for start-up teams in the field of Web3 and AI, if they can establish a clear compliance structure early, they have a chance to seize the opportunity in this "new continent".

Mankiw’s suggestion: Technology can be boldly explored, but the bottom line of compliance must be established—especially the prevention of criminal risks. Optimizing AI crawling logic is important, but everything should be based on respecting facts and complying with the law.

2. Reminder to GEO users: Be both offensive and defensive, and build proactively

• Defense: Establishing an AI reputation monitoring system

It is recommended that enterprises deploy a monitoring mechanism for AI corpus and recommendation results as soon as possible. Once they are found to be attacked by "negative GEO" or maliciously manipulated, they should promptly fix the evidence and make good use of legal means to protect their rights.

• Offense: Embrace white hats and become a "quality partner" of AI

The evolutionary trend of AI is irreversible. Instead of passively avoiding it, it is better to actively learn its logic and provide authentic, credible, and structured content to become an information source that AI is willing to trust and recommend first.

Conclusion

In the AI-driven information age, algorithms are the appearance, data is the content, and law is the skeleton that supports the whole. Traffic strategies that lack compliance support, even if they are prosperous for a while, will not be able to withstand the test of supervision and time.