AI has spurred the emergence of yet another trillion-dollar market.

Recently, the "Baidu GEO Marketing Plan" has been making rounds within the AI community, sparking curiosity among numerous netizens: "Is Baidu venturing into GEO?"

GEO (Generative Engine Optimization) entails optimizing webpage content to increase the likelihood of brand information being referenced in responses from AI chat assistants (such as Doubao, Qianwen, DeepSeek, etc.). As the user base and usage of AI chat assistants expand, brands are also setting their sights on GEO marketing.

According to a report by "LeadLeo Research Institute," the global GEO market is projected to reach RMB 373.9 billion by 2029, with a five-year compound annual growth rate of 139.2%. Traditional search engines once dictated a generation of online marketing strategies. Now, with traffic shifting towards AI applications, will major AI players like ByteDance and Baidu step in to redefine marketing rules in the AI era? Will the extensive content deployment during the GEO process lead to "information pollution," akin to the early days of SEO?

01. Brands Flock to GEO

Recently, a PowerPoint presentation titled "GEO Optimization Solutions," purportedly from Baidu Marketing, has been circulating on social media. The PPT outlines the value of GEO and how brands can enhance their visibility in AI assistants through GEO, leading some netizens to speculate that Baidu is entering the GEO market.

A source close to Baidu clarified that the plan was not officially produced by Baidu Marketing, but acknowledged that Baidu Marketing does indeed have plans to engage in GEO, which are currently under exploration.

Over the past two decades, SEO has been the cornerstone of online marketing. Brands vied for top positions on search engine result pages through keyword placement, backlink building, and content stuffing. However, this paradigm is now being disrupted by AI search.

AI-generated answers are influencing people's decision-making processes. Many users now turn to AI to "screen" options before making purchases, such as finding face masks suitable for sensitive skin or fans suitable for office desks, preferring AI recommendations over manual keyword searches.

Data from "LeadLeo Research Institute" reveals that the monthly active users of large model applications globally have surpassed 900 million, with 85% of them possessing information retrieval capabilities. Over 80% of users have quickly developed a high daily dependency. Tan Beiping, Dean of the Miaozhen Marketing Science Academy, suggested that user behavior habits have undergone a fundamental shift, characterized by "clickless searches," where users no longer click on blue links below, potentially leading to a 60% or more decline in traditional search engine optimization business.

The shift in user search behavior has left many brands anxious. With the explosive popularity of DeepSeek and the surge in daily active users of Doubao, many brands have ramped up their demand for exposure on AI platforms this year. Industry insiders bluntly state that almost all market departments of national brands are discussing how to implement GEO. Zhao Jie, founder of "Yanlin Cloud Innovation," a company specializing in GEO services, noted that many brands hope users will see and understand their products, especially hoping negative public opinions about the brand will not appear in AI.

Similar to SEO, both SEO and GEO aim to secure prime positions in search engines; the top spots in traditional search results and the brands or enterprises prioritized in AI responses are both considered highly valuable. Lulu (pseudonym), involved in GEO services, stated that many brands fear missing out and worry that once these prime spots are occupied, GEO will follow SEO into a bidding war, increasing brand layout costs. "Many brands fear being outperformed by competitors and are jumping on the GEO bandwagon, further heating up the market."

Some GEO service providers claim that for just RMB 20,000, they can potentially get a brand featured in the Q&A sections of three mainstream AI platforms and maintain visibility for three months. "Qujie Business" learned that GEO is still in its infancy, with significant variations in pricing and standards among different service providers. Zhize (pseudonym), with years of experience in the SEO industry, stated that his company has also expanded into GEO services, typically charging based on "keywords" and "AI platforms," with the monthly purchase price for a single keyword usually ranging from RMB 2,000 to RMB 5,000. If a brand purchases 20 keywords and a portion is included by AI, it is considered meeting the standard.

02. Commercialization of AI Responses?

Zhao Jie, founder of Yanlin Cloud Innovation, who previously served as president of a leading maternal and child group, believes that currently, during this user migration transition period, GEO has not yet formed the traditional SEO-style bidding rankings, and the algorithm logic and reasoning preferences behind large models are constantly evolving, presenting numerous opportunities for refined service operations.

The booming GEO market has attracted a growing number of players. Some are traditional SEO service companies transitioning into GEO; others are "tech-savvy" startups leveraging self-developed tools to enhance content optimization efficiency. There are also GEO service companies founded by entrepreneurs with extensive marketing or technical experience, who, although lacking self-developed models, possess service delivery and client mining capabilities.

Different types of companies employ varying strategies. SEO enhances search rankings by improving webpage content quality and keyword relevance; GEO, on the other hand, builds upon quality and relevance to create content more easily referenced by AI.

Some GEO service providers transitioning from SEO possess abundant social media account and channel resources, implementing GEO through extensive content placement or even machine-generated content, albeit with relatively crude results. In contrast, tool-based service providers with a technological focus emphasize understanding the underlying logic of large models and systematically designing content strategies.

Yanlin Cloud Innovation, where Zhao Jie works, first converts brand materials into document formats suitable for AI training, then trains an "article generation agent" capable of producing content that complies with AI's webpage information retrieval rules. After generating multiple articles with human assistance, they are distributed through channels easily referenced by AI, such as knowledge bases, to increase the likelihood of being referenced in AI responses.

Among domestic tool-based GEO service providers, PureblueAI has garnered significant attention. Founded by Lu Yang, the former market leader of Doubao's large model, the company secured multi-million yuan seed funding in September, co-led by BlueFocus and InnoAngel Fund.

According to "36Kr," Qinglan Intelligence has developed an end-to-end model and Agent product that learns the recommendation patterns of AI platforms, continuously analyzing content successfully recommended in AI searches to identify common "feature factors." When a client proposes optimization needs (e.g., optimizing the search intent for "recommended super hybrid architecture vehicle models"), Qinglan's model not only generates a marketing draft but also calculates the most effective platform combination for publication.

Besides various service providers, whether the platforms behind AI chat assistants will engage in GEO has also drawn significant attention from both inside and outside the industry.

Overseas, the popular AI search engine Perplexity has started recommending advertisements in the "Related" section below its answers, marked with a "sponsored" label.

Many industry insiders believe that domestic AI chat assistants will not rely on advertising for monetization until they have firmly captured their user base. However, once user scales reach a certain level, commercialization is likely to follow, with a higher probability of introducing advertising slots for brands. Nevertheless, they are unlikely to collectively engage in GEO content deployment. Yu Jian, co-founder of Yuanyi Information, previously stated that platforms will inevitably enter brand marketing in the future, offering B-end brands more imaginative business models beyond simple bidding rankings, such as providing comprehensive "solutions" capable of influencing user decisions.

03. GEO Still Navigating Uncharted Waters

Alongside the surging market enthusiasm, GEO has also sparked controversy.

Due to the diverse backgrounds and varying levels of competence among numerous service providers, brands often struggle to find suitable partners. Even when they do, some providers cannot guarantee specific results in terms of traffic generation or conversion.

This stems primarily from the complete invisibility of large models' retrieval and reasoning processes, also known as the "black box" nature of large models. Service providers can lay out content according to AI's preferences for information retrieval, such as analytical articles with clear headings and structured layers. However, numerous factors influence the answers when users pose specific questions, including the timing of the question, IP address, and previous dialogue content, all affecting the generated results.

Meanwhile, mainstream large models undergo major updates every 1-3 months, each potentially retraining or adjusting reference preferences, rendering previously deployed content by service providers ineffective. For this reason, the current delivery cycle for GEO services is generally settled on a monthly or quarterly basis.

Lulu stated that overseas GEO practitioners can use third-party tool platforms like "semrush" and "Profound" to track which website traffic comes from AI and which from organic sources, seeing whether published GEO-optimized articles have been successfully referenced and which are referenced the most. However, in China, with large model data from DeepSeek and others not openly available, no third-party tool can yet process data across all AI platforms, making effectiveness harder to quantify.

"Qujie Business" learned that many brands currently engage in GEO primarily to enhance exposure and optimize public sentiment. Some service providers noted, "Brands hope the information AI references about them is not outdated and that their exposure rate is not lower than competitors. However, achieving exposure and specific sales conversions is more challenging."

Concerningly, the "black box" nature of large models has also given rise to numerous gray or even non-compliant operations. Some eager service providers have resorted to fabricating authoritative sources, registering high-authority self-media accounts in bulk, and flooding AI with low-quality content. Such "feeding" may introduce false information, exaggerated claims, or even malicious smears into AI responses, forming "information pollution."

Recently, tech blogger "Digital Life Kha'Zix" conducted a test to deceive AI, fabricating notes about himself on platforms like Toutiao and Sohu, successfully "polluting" AI and making himself appear as the son of "Hajime" and "Northeast Rain Sister" in DeepSeek's responses.

GEO is becoming a new battleground for marketing, yet it remains a vacuum zone lacking transparency and regulation. "Qujie Business" noticed that some companies are already advocating for influencing AI through high-quality content rather than technical loopholes, establishing a detection system for GEO indices to promote a healthy GEO ecosystem.

As platform mechanisms gradually become transparent and third-party monitoring tools mature, GEO is expected to transition from "wild growth" to "refined operations."