The new "Yi Zhongtian" emerges! Rewriting the underlying logic of advertising marketing in the AI era: GEO

Today, the AI application sector in the A-shares market has welcomed a long-awaited collective explosion. On the market, stocks such as ZhiDeMai and ZhuoYi Information hit the daily limit; ECLICKTECH, COL, and TLJT formed a new "Yi Zhongtian" combination, all recording a 20CM limit up in the morning. At the same time, BlueFocus surged to the daily limit with a trading volume of 19.32 billion yuan, ranking first in the A-shares trading volume for the day. Beneath the surface heat, the industry is undergoing a more profound disruptive transformation— as user decisions shift from "clicking links" to "reading AI-generated answers," the logic of brand visibility is being completely rewritten. A brand new competitive approach is emerging: GEO (Generative Engine Optimization) is becoming the traffic entry point and marketing survival rule in the AI search era. As users no longer "click links," the foundation of the advertising industry begins to loosen Imagine this scene: At eleven o'clock at night, you open your phone, just wanting to casually ask, "Which baby formula is suitable for sensitive stomachs?" In the past, you would open a search engine, look at a string of blue links, jump to reviews, browse e-commerce, and then make a judgment amidst conflicting opinions. But this time is different. You open AI search: it directly gives you a "doctor-like" conclusion, clearly organized, explained in points, and conveniently lists the pros and cons of several products along with their suitable user groups. You don’t even need to click on any webpage—the answer is already "complete" in front of you. A chart from CITIC Securities illustrates this change more straightforwardly: traditional search is "search—browse list—click link," while AI search is more like "search—directly read answer," with links being postponed or even skipped. At this moment, the foundation of marketing begins to loosen. For the past twenty years, a brand's "visibility" on the internet has heavily relied on one action: user clicks. SEO, bidding rankings, information flow placement, content seeding... no matter how the methods change, the essence is still competing for that one click, bringing people into your page, your detail page, your store. But AI search has folded the process: Users no longer need to "enter a webpage" to obtain information. Information is read, digested, and reorganized by AI, then delivered directly in the form of answers. Thus, a brand new question emerges: CITIC Securities presented this migration in a chart: Past: Search → Browse link list → Click webpage Now: Search → Directly read AI-generated answers (links can be completely skipped) When links are no longer entry points, the brand exposure mechanism is completely changed. How can brands be "seen" when users no longer click on links? How can brands be "actively mentioned" when the answers are generated by AI? This is the root of the GEO explosion—not a marketing trick, but a survival strategy after the migration of entry points. 1. Migration of Entry Points: Advertising budgets always chase "time" The essence of advertising is to reach more audiences; where the audience's time is, the budget will flow there. When links are no longer entry points, the brand exposure mechanism is completely changed. Guotai Junan reviewed a clear migration chain in their research report: In the PC internet era, search engines and portals were the entry points; In the mobile internet era, super apps and information streams/short videos became the entry points; In the AI era, entry points further migrate to AI platforms like DeepSeek, Doubao, Kimi, etc., with the "search box" transforming from an entry to a webpage list into an interface for "instant knowledge synthesis." This migration is not something that "may happen in the future," but is currently happening. CITIC Securities cited data showing that ChatGPT and Doubao App have monthly active users of 780 million and 170 million, respectively, with penetration rates exceeding 10%; while traditional search entry points like Baidu fluctuate between 650 million and 750 million monthly active users, trending towards stability. More importantly, consulting agencies have provided more aggressive judgments: Gartner predicts that by 2026, search engine traffic will decline by 25%, and the share of search marketing will be taken over by AI chatbots and virtual agents. This means: The certainty logic of "pay-per-click" in search advertising is being diluted; the competition for "answer exposure" will become the new main battlefield. 2. Why SEO suddenly doesn't work: Click-through rates "collapse" in front of AI answers When AI starts to provide answers directly on search result pages, users naturally click less. CITIC Securities cited observations from Seer Interactive: When Google’s AI Overviews are triggered, the natural click-through rate drops from about 1.5% to about 0.5%, and the paid click-through rate also shows a downward trend. The harsh reality of this data is: It is not a "fluctuation in a certain industry," but a structural loss You may have optimized your website to the first or second position, but you might still have "won the ranking" and lost clicks; You may have spent money to buy a higher position, but you might still have "bought exposure" and not gained visits; Because users end their decision-making on the results page—without even giving you a chance to be "glanced at." Thus, the classic slogan of SEO—"make it easier for users to find you"—is quietly being rewritten as: Make AI more willing to mention you. Make AI say you when answering. III. What exactly is GEO: From "being searched" to "being mentioned by AI" CITIC Securities defines GEO very precisely: GEO is a type of advertising marketing technology service, with the core being to make brands actively mentioned in AI searches. If SEO optimizes for "ranking," then GEO optimizes for two more hidden and fundamental aspects: The recognition of brand content by large models The credibility of brand content by large models Guotai Junan describes the difference in more "technical" terms: the core mechanism of traditional search is matching (inverted indexing, PageRank), and SEO is deterministic optimization; the core mechanism of AI search is analysis (vector retrieval + RAG), and GEO is probabilistic optimization—every action you take increases the probability of "being recalled, being trusted, being cited." This is also why GEO seems like marketing, but at its core is more like "content engineering + trust engineering." IV. How AI search generates answers: Understanding RAG to grasp the leverage point of GEO Many people think that AI search is just "the model says whatever it remembers." However, Guotai Junan breaks down the process in detail in their research report: Currently, AI search commonly adopts the RAG (Retrieval-Augmented Generation) architecture. You can think of it as an assembly line: Step 1: Storage Web content is cut into multiple chunks, transformed into vectors using an embedding model, and stored in a vector database. Here, content is not "text," but "coordinates." Step 2: Semantic Retrieval User questions are also vectorized, and then the system calculates similarity to recall the Top-K relevant chunks. Here, AI is not matching keywords, but matching "intent." Step 3: Context Injection and Generation The recalled chunks are fed into the model's prompt, where the model uses attention mechanisms to determine which are more credible and informative, and then generates the final answer Here, "clear structure, clear entities, and containing data" chunks are easier to be cited. Thus, you will find that the key points that GEO can leverage happen to fall on two stages of this production line: Information Retrieval: Making your content easier to be recalled by vector databases; Content Evaluation: Making the model more willing to treat your content as a "source of facts." This is why GEO is not about "writing more keywords," but rather "more like writing a manual for the model." V. How GEO Works: More Content is Not Better, but More "Citable" is Better The GEO paper cited by CITIC Securities, "GEO: Generative Engine Optimization," lists seven commonly used optimization methods and conducts experimental comparisons. One of the most effective methods is "adding direct quotes from celebrities or institutions closely related to the topic," which can bring about a 40% increase in exposure; "adding specific statistical figures" can also lead to significant improvements. Behind this actually reveals a simple rule: Models prefer expressions that can be treated as evidence. If you write "Our sales are great," the model won't take it seriously; If you write "According to statistics from a certain institution, the Chinese GEO market size in Q2 2025 will grow by over 200% year-on-year," the model is more willing to cite it. Guotai Junan further proposed a framework that resembles a methodology: the DDS principle— Semantic Depth Data Support Authoritative Source These three together construct "content that is prioritized by AI for citation." If we translate DDS into more relatable terms, it would be: Be clear, be thorough, be supported by evidence, and sound real (and preferably actually be real). VI. Platform Preference: Where You Write Determines Who Sees You Many brands are accustomed to writing content "on their own official website." However, AI's knowledge sources do not only look at official websites. CITIC Securities cited research from the overseas GEO company Profound: the platforms most cited by ChatGPT include Wikipedia, Reddit, Forbes, etc.; Google AI Overviews and Perplexity prefer Reddit, YouTube, Quora, Gartner, etc. The same is true domestically. CITIC Securities took "baby formula recommendations" as an example and analyzed the citation sources of DeepSeek, Doubao, and Yuanbao: the three major platforms clearly prefer maternal and infant vertical media and comprehensive portals; among them, Yuanbao has a significant citation proportion for "WeChat public accounts." This means that GEO is not just about "good content writing," but also about "correct content publishing." You need to enter the model's "high-frequency trust channels" to be more likely to appear in the answers. VII. The business model is changing: Advertising companies have the opportunity to make money like SaaS companies for the first time This may be the most worth reading section repeatedly by "industry people" in the two research reports: GEO is not only a new advertising method but may also drive advertising agencies from "human-based services" to "technology-based services." CITIC Securities mentioned that GEO companies generally adopt monthly subscriptions or project-based models. Taking the overseas star company Profound as an example, the monthly subscription is $399, which can track 100 prompts for one brand and three AI search products; there are also customized services. Other institutions quote about $30,000 to $200,000 per month, tied to KPIs (such as ChatGPT Top-3 citation rate, Perplexity citation rank, etc.). Guotai Junan also provided more complete "tiered pricing" information: Profound's startup version is $99/month (only tracks ChatGPT, with a limit of 50 prompts), the growth version is $399/month (monitors three AI platforms, 100 prompts, and six optimized articles per month), and the enterprise version is customized; it also mentioned that its client base has reached about 500 corporate clients. Looking at this string of numbers, you will find that GEO's charging model resembles software: Subscription-based Per seat/Quota-based Based on the number of tracked prompts/platforms Tied to quantifiable metrics And this is precisely what traditional advertising agencies find most difficult to achieve but most desire to achieve: Transforming from one-time project fees to recurring "recurring revenue." Why is the industry excited? Because the profit structure will change. CITIC Securities provided a comparison: The SEO industry has been developing for nearly 30 years and is still extremely fragmented, with the leading Semrush's revenue scale being about $400 million and a market share of about 0.5%. What does fragmentation mean? It means low barriers, low premiums, and relying on a manpower strategy. But GEO has higher barriers: The black box and randomness of large models Generation differences across platforms The process of "making the model understand and cite" is more complex Therefore, market concentration may actually increase Once the concentration increases, the leading players have the opportunity to achieve "software-like" economies of scale: More data → Better model evaluation → Stronger optimization capability → Higher renewal rates → Stronger data feedback loop. 8. Market Space: Why "Ten Billion Dollars is Not a Story," but an Inevitable Result Benchmarking SEO Two research reports provide very straightforward "anchors" for the market space. CITIC Construction Investment: The global SEO market size is expected to be around $80 billion in 2024, and GEO is expected to replace traditional SEO in the AI era, with the market size likely reaching the ten billion dollar level. Guotai Junan goes further to provide pathways and numbers: The global GEO market size is expected to be around $11.2 billion in 2025, with China around 2.9 billion; It is estimated that by 2030, the global market will exceed $100 billion, with China around 24 billion (first page of the report "Introduction to this Report"). The logic behind these predictions is not complicated: Migration of entry points → Migration of attention → Migration of budgets → Outburst of new optimization services. Moreover, GEO has a "window of opportunity": Guotai Junan believes that the advertising monetization ratio of large models in the short to medium term is still low, creating a "monetization vacuum period," allowing GEO companies to establish "brand visibility management" as a business before the platform fully consolidates commercialization. This is reminiscent of the early information flow of the mobile internet: when the platform's advertising system is not mature, service providers often move the fastest. 9. If You Are a Brand Owner: A More Realistic GEO Startup Checklist Many articles discussing GEO stop at "the concept is very hot." However, CITIC Construction Investment and Guotai Junan have actually laid out the execution path very clearly. CITIC Construction Investment provides a six-step method: Intent analysis: Sort out how users will ask you Information sorting: Inventory all public/internal materials you can "feed to AI" Content structuring: Transform long articles into Q&A, data lists, and other semi-structured content Semantic optimization and authoritative endorsement: Precise expression + expert/institution endorsement Multi-modal and multi-platform adaptation: Rewrite and distribute according to platform style Continuous monitoring and iteration: Track mention rates/positions/emotional tendencies and optimize in cycles Guotai Junan leans more towards "technical actions": A clear Header structure affects Chunk slicing boundaries Clearly defined entities (brands, products, experts) can enhance vector retrieval features Inverted pyramid writing can counteract context window limitations and attention decay Combining the two research reports into a "practical" version, the key points can be summarized into four keywords: Prompt Assets, structured knowledge base, authoritative distribution, visibility monitoring. You are not "producing more content," but rather building a set of "brand knowledge infrastructure for the AI era." 10. Final Reminder: The ceiling for GEO is high, but the pitfalls are deep Any "optimization" will lead to gray industries, which is a historical pattern. In the early days of SEO, it was about keyword stuffing and backlink manipulation; GEO will inevitably see "pseudo-authoritative endorsements," "data collages," and "content factories." Moreover, AI inherently has black box issues, randomness, and platform differences, and both research reports list these as challenges: opaque citation logic, significant differences between AI platforms, and results with random attributes. Therefore, the true long-term players in GEO may not be those who are best at "exploiting loopholes," but rather those organizations that can continuously produce high-quality, verifiable, and traceable content assets. As AI becomes the new "medium," brand competition will return to an old but often overlooked proposition: Who is more trustworthy will be more easily repeated. Who is more easily repeated will be more easily chosen. Conclusion: The next generation of advertising may no longer be about "buying exposure," but rather "being written into answers" If we were to compress the history of internet advertising into one sentence: In the PC era, it was about ranking; in the mobile era, it was about recommendations; in the AI era, it is about citations. The emergence of GEO essentially reminds every brand: You need to switch from "traffic thinking" to "answer thinking"; From "pulling people in" to "letting AI speak for you." As users become increasingly accustomed to asking AI for conclusions in a single sentence, the fate of brands will increasingly depend on: Whether you are mentioned in that conclusion; how you are mentioned; and whether you are considered a credible option. This is not a minor adjustment for the marketing department, but a system upgrade following a migration of entry points. You may not like it, but it is hard to avoid it