Today, artificial intelligence (ai) is increasingly present in people's daily lives, and many people are becoming a habit of “unquestioned ai”. Recently, however, a number of users have found that an ai response that would have been objective was actually hidden in advertising. This new type of advertising that uses the ai network's “capturing” logic, with “private goods”, has led to the emergence of an emerging industry, commonly known as geo (generative engine optimization, general engineering)。
Geo is the product of the deep integration of digital marketing and ai technology, which has emerged mainly from four factors: the shift in user behaviour towards ai, the upgrading of technology architecture, changes in market demand and the optimization of traditional search engines (seos) has been gradual in the ai era. Since its emergence in 2023, geo has become a key strategy for businesses to compete for flows during the ai era。
The china internet information centre (cnnic) report on the development of generating artificial intelligence applications (2025) shows that, as of june 2025, china's generation of ai users had exceeded 515 million, and the generation of ai has been widely used in intelligent search, content creation, etc. With the spread of aai mobile phones and primary ai hardware, the traffic portal for questions and searches has been completely turned to ai. According to the 56th statistical report on the development of the internet in china, by june 2025, the size of our search engine was 69. 2 per cent of the total number of internet users, a significant decline from 78 per cent, 75. 7 per cent, 75 per cent and 79. 2 per cent of the previous four surveys。
Such changes in the market environment have brought “east wind” to the development of geo. The data show that the domestic geo market was over 4. 2 billion yuan in 2025, with a compound growth rate of 38 per cent in the last three years. As more and more enterprises focus on geo marketing, it becomes increasingly important to distinguish between the objective learning results of ai and advertising content。
In terms of the technical essence, geo has the rag (retrieval enhancement generation) architecture at its core, building a full-link system from “content production” to “ai acceptance” through the three pillars of vector databases, dynamic knowledge mapping and multi-modular adaptation. In short, it is by adjusting the structure and semantics of the content published online to make it easier for it to be retrieved by ai, quoted and generated into the final answer, thereby influencing the information that users see in the ai interface。
This pattern significantly reduces traditional information access and decision-making links. According to statistics, the average time spent by users reading ai-generated answers was only 8. 7 seconds, a 62 per cent decline from the traditional page view. In the process, the user's trust mechanism has also been migrated - – 78 per cent of users believe that the “authority sources” of ai integration are more credible than individual websites. This is also being used by many geo practitioners to guide ai in generating responses that meet specific business objectives。
There are two typical “optimal” instruments. The first is semantic quantification. Large models typically transform text into high-dimensional vectors, generating answers by calculating similarities and probability distributions between vectors. Geo is able to improve the probability of being quoted by adjusting content to bring target information closer to the user's query word in the vector space. Some of the practitioners are also targeted, have a large number of duplicates or low-quality content for ai capture. This is essentially a “data pollution” act。
The second is to enhance the authority and credibility of content by various means. At the request of the platform, ai would prefer to refer to a clear structure and authoritative source. Geo has defrauded the credibility of the larger model through additional data supporting, citing sources, and sometimes even making up “authority reports” or using “false experts” to “false science generals”。
The disorderly development of geo must be accompanied by a series of legal and ethical challenges, resulting in ethical conflicts between business interests and information neutrality, technological manipulation and ecological fairness. These conflicts are often undetectable, and they can be lurking and deepening over time. Ultimately, it may have a profound impact on the quality of the information we are exposed to, on the users ' own judgement and interests, and on a healthy market competitive order。
There is an urgent need to promote the healthy development of the geo industry in three areas: first, the adoption and purification of language banks. Large-scale model manufacturers should identify and clean contaminated content in a timely manner through technical means after capturing the web language. Industry organizations are also required to provide criteria for defining “speech pollution” at the substantive level, while supporting industry development to deter malicious data manipulation。
The second is to guard against increased risk in highly sensitive areas. Particularly in areas such as health, finance and public policy, where the authenticity of information is extremely demanding, geo, if abused, can directly cause real harm. This requires constant refinement based on existing laws and regulations in order to protect public safety and interests through stricter regulation and accountability。
Thirdly, there is a clearer distinction between advertising and ordinary content. Our current law on advertising clearly states that “advertisements should be identifiable”. On this basis, the generator engine should be required to clearly label “advertise” or a similar hint when returning to the geo to adjust content to avoid misunderstanding by the user. Failure to comply with the labelling obligation is subject to prosecution under the corresponding legislation, thus forcing the industry to regulate itself。
In the long term, the industry also needs to focus on deeper issues such as user privacy and cognitive sovereignty, monopoly of knowledge resources and protection of content property rights. This requires continuous communication between the academic community and the industry to develop a clearer consensus and to further promote the landing of corresponding norms and measures. Only timely external regulation and self-regulation within the industry can safeguard the gradual release of the potential of emerging industries such as ai. (the author is the vice-president of the national council for innovation and development strategies of china)
