Zheshang: The search logic has undergone a fundamental change, highlighting the importance of authoritative information sources.

Zheshang report stated that the illusion problem of large models is inherent in the underlying technology path of artificial intelligence, and the problem of the surge of false information in AI awaits resolution. The bottleneck period of base large model pre-training has been reached, and there is an urgent need to solve the difficulty of sustainable supply of high-quality data. At the current time, from SEO to GEO, the logic of search has undergone a fundamental shift, and the Internet is transitioning from the "ranking grabbing" SEO era to the "answer grabbing" GEO era. The importance of authoritative information, in both the training data sources of large models and from a commercial perspective, has significantly increased. The main points of Zheshang are as follows: Information overload and quality crisis At the current time, the surge in false information in AI has led to phenomena such as machine illusions, information overload, and information pollution, causing a trust crisis in the application and development of large models. Developers of large models in various countries have increased their research and development efforts, attempting to enhance the credibility of basic large models through larger scale parameters and more training data. However, the illusion problem of large models is inherent in the underlying technology path of artificial intelligence. Currently, the pre-training of base large models has reached a bottleneck period, and there is an urgent need to solve the difficulty of sustainable supply of high-quality data. Research company EpochAI estimates that by 2028, all high-quality text data on the Internet will have been used up. Possible solutions include the introduction of data from authoritative knowledge suppliers, including from book publishing organizations, broadcasting organizations, libraries, archives, etc., and the inheritance and upgrading of traditional knowledge consensus mechanisms. At the current time, the importance of authoritative information sources in large model training is rapidly increasing. At the current time, from SEO to GEO, the logic of search has undergone a fundamental shift, and the Internet is transitioning from the "ranking grabbing" SEO era to the "answer grabbing" GEO era. The importance of authoritative information, such as various official media, has significantly increased in both the training data sources of large models and from a commercial perspective. Specifically, the content of authoritative media has become the basic material for constructing the knowledge system of large models, determining the positions of information nodes and the strength of relationships. The Beijing Nanwen Technology team's analysis found that traditional media content has a weight of up to 73% in credibility assessments, far exceeding that of self-media content. From being "squeezed out" to win-win cooperation, authoritative media provide trustworthy content, while technology platforms provide technology and traffic, forming a new value ecosystem. Currently, many large model enterprises are actively seeking cooperation with authoritative media to reduce illusion risks and improve content quality. The commercialization effects of cooperation with authoritative media are also expected to improve. From the perspective of "answering", the depth of reporting by authoritative media (such as People's Daily and Xinhua News Agency) is labeled as important in AI knowledge graphs. In the training mechanism of deep learning models, reports from Xinhua News Agency, "Caijing" magazine, and other media are often labeled as high-weight sources of credibility. This technical feature makes traditional media content the "structural beam" of the knowledge graph, determining not only the positions of information nodes but also influencing the strength of relationships. According to data from Beijing Nanwen, the penetration rate of company information cited by three or more authoritative media sources in AI-generated content is increased by 80%. Regarding targets At the current time, it is recommended to focus on authoritative information publishing institutions such as People.cn Co., Ltd. (603000.SH), Xinhuanet Co., Ltd. (603888.SH), Sichuan Newsnet Media (300987.SZ), Southern Publishing and Media (601900.SH), China Publishing & Media Holdings (601949.SH), Oriental Pearl Group (600637.SH), etc. Risk alert The development progress of large models is lower than expected, and changes in international trade policies, etc.