China: AI Manipulated by GEO to Spread Misinformation

GEO: An AI Manipulation Strategy in China

In China, GEO services, which are sold on platforms like Taobao and JD.com, have attracted attention due to their ability to manipulate search engine results based on artificial intelligence. Subscriptions to these services vary significantly, ranging from 3,600 yuan (approximately $520) to 32,800 yuan (approximately $4,765), according to reports from local media.

This phenomenon gained momentum last year, coinciding with the integration of AI into search platforms. Mr. Wang, the director of a major GEO service company, revealed that his firm had served over 200 clients in a year. His company helped these clients rank among the top three results on various AI platforms.

The Mechanics Behind GEO

According to Mr. Wang, the success of this strategy relies on the continuous provision of large amounts of content to AI models such as DeepSeek, Doubao, and Kimi. This content is then analyzed, indexed, and highlighted in search results, thereby influencing the responses provided to users.

GEO was originally designed to enhance the visibility of brands and their products in AI-based searches. However, some Chinese companies exploit this technology to manipulate AI with misleading information that exaggerates their products. This practice was highlighted on March 15 during the Consumer Rights Gala organized by CCTV.

How GEO Services Operate

Li, the founder of Liqing GEO, explains that the core of their service is based on publishing optimized advertisements and press releases on specific websites. These sites, once unprofitable, are now flooded with orders, sometimes publishing hundreds of articles per day.

To illustrate the process, an insider purchased the "Liqing GEO Optimization System" software and created a fictitious product, a watch called "Apollo-9." Using the software, he automatically generated more than a dozen promotional articles that were published on social media accounts. In just two hours, a major AI model used this content to respond to a question about the value of the Apollo-9 watch, even recommending it for its features. The insider then continued to publish eleven additional articles in the following three days, including fake expert reviews and rankings.

The Apollo-9, although non-existent, thus appeared among the top suggestions from AI models for smartwatches, illustrating the effectiveness of this manipulation.

Implications for Users and Brands

For users, AI is supposed to provide clear and reliable answers. However, with these distorted GEO practices, trust in the information provided by AI is undermined. Behind an apparently relevant response may lie artificial content designed to influence rather than inform.

Users must remain vigilant and not place blind trust in the answers provided by AI, as machines can make mistakes or "hallucinate" information. Platforms like ChatGPT warn against these errors and recommend verifying important information.

For honest brands, the situation is frustrating. Those that focus on quality and transparency find themselves overwhelmed by a flood of overly optimized content, creating an unfair competition where visibility depends on the ability to produce massive amounts of content, even at the expense of truthfulness.

Since the gala, several GEO companies have denounced the "brainwashing" of AI and promised to reduce misinformation in their services, although the effectiveness of these promises remains to be seen.

Towards Stricter Regulation

In light of these abuses, stricter regulation seems inevitable. Beijing has already demonstrated its ability to intervene quickly on sensitive digital issues. For example, the law firm White & Case reports that by 2025, Chinese authorities imposed mandatory labeling of AI-generated content, with strict rules for platforms to limit abuses and strengthen information control.

While specific rules regarding GEO services have not yet been published, AI platforms may also be required to bolster their defenses. Currently, these models analyze and synthesize large amounts of content, but they still struggle to distinguish reliable information from artificially amplified content.

It is therefore likely that more sophisticated filtering systems will be developed to detect manipulation campaigns or suspicious postings. Furthermore, there is a risk that actors may manipulate AI not only to promote their products but also to harm the visibility of a competitor, a practice known as "Black Hat."