生成式引擎优化（GEO）团体标准如何应对AI虚假信息风险？2026年治理规范解读

引言

随着生成式人工智能（GAI）开始重塑信息获取的方式，一种新的生成式引擎优化（GEO）A content optimization strategy for generative AI search engines, focusing on making content understandable and recommendable by AI systems.赛道正迅速崛起。在AI颠覆传统迎来新机遇的同时，“语料投毒在AI训练数据或可抓取内容中恶意植入虚假、伪造或负面信息的行为，旨在污染AI信息生态。”、虚假信息、算法偏见等风险也浮出水面，各界对“信任”的要求更显突出。4月10日，一项旨在为这一新兴领域立规矩、划红线的团体标准《生成式引擎优化（GEO）A content optimization strategy for generative AI search engines, focusing on making content understandable and recommendable by AI systems.可信信息传播与信息生态治理规范》（以下简称《规范》）正式通过立项评审，标志着我国在人工智能信息生态治理领域迈出了关键一步。

As Generative Artificial Intelligence (GAI) begins to reshape how information is accessed, a new field known as Generative Engine Optimization (GEO) is rapidly emerging. While AI disruption brings new opportunities, risks such as "corpus poisoning," misinformation, and algorithmic bias have also surfaced, making the demand for "trust" more prominent across all sectors. On April 10th, a group standard titled "Specification for Trusted Information Dissemination and Information Ecosystem Governance in Generative Engine Optimization (GEO)" (hereinafter referred to as the "Specification") officially passed its project approval review. This marks a crucial step for China in the governance of the AI information ecosystem.

该标准由中国新闻技术工作者联合会归口，新华网融媒体未来研究院、新华社国家重点实验室生物感知智能应用研究部、新华社中国广告联合有限公司、新华社通信技术局、复旦大学马克思主义研究院人工智能伦理与意识形态安全实验室、北京邮电大学网络安全学院、新经济学家智库、江苏天闳低空数字技术研究院等多家权威科研与产业机构共同发起。即日起向全国征集起草人与起草机构献计献策、共同参与。

The standard is administered by the China Association of News Technology Workers and jointly initiated by several authoritative research and industrial institutions, including Xinhuanet Media Future Research Institute, the Bioperception Intelligent Application Research Department of Xinhua's State Key Laboratory, Xinhua China Advertising United Co., Ltd., Xinhua Communications Technology Bureau, the AI Ethics and Ideological Security Laboratory at Fudan University's Marxism Research Institute, the School of Cyberspace Security at Beijing University of Posts and Telecommunications, the New Economist Think Tank, and Jiangsu Tianhong Low-Altitude Digital Technology Research Institute. From today, it is soliciting contributions and participation from drafters and drafting institutions nationwide.

GEO的兴起与挑战

随着ChatGPT、DeepSeek等大模型应用的普及，用户正习惯于向AI提问获取答案。相应地，品牌方也开始通过优化自身内容，以期在AI生成的回答中获得更高的“可见度”，这便是GEO的雏形。但问题也随之而来：虚构专家资质、伪造行业数据、恶意植入竞品负面信息等“语料投毒在AI训练数据或可抓取内容中恶意植入虚假、伪造或负面信息的行为，旨在污染AI信息生态。”行为时有发生，一旦被大模型抓取并反复引用，就会在AI信息生态中形成“持续性污染”，其负面影响已开始涌现。

With the widespread adoption of large model applications like ChatGPT and DeepSeek, users are becoming accustomed to obtaining answers by asking AI. Correspondingly, brands have begun optimizing their own content to achieve higher "visibility" in AI-generated responses, which is the nascent form of GEO. However, problems have emerged alongside this trend: "corpus poisoning" behaviors such as fabricating expert credentials, falsifying industry data, and maliciously inserting negative information about competitors occur from time to time. Once captured and repeatedly cited by large models, these can create "persistent pollution" within the AI information ecosystem, the negative impacts of which have already begun to surface.

“技术的进步不应以牺牲信息的真实性为代价。”中国新闻技术工作者联合会相关专家表示，“当前市场上，GEO服务良莠不齐，部分行为已游走在虚假宣传与不正当竞争的灰色地带。这份规范的制定，就是要在行业发展初期就植入‘安全’与‘可信’的基因，避免重蹈互联网早期信息失序的覆辙。”

"Technological progress should not come at the cost of information authenticity," stated a relevant expert from the China Association of News Technology Workers. "Currently, the quality of GEO services in the market is uneven, with some practices operating in the gray area between false advertising and unfair competition. The formulation of this specification aims to embed the genes of 'security' and 'trustworthiness' into the industry at its early stage of development, preventing a repeat of the information disorder seen in the early days of the internet."

《规范》核心治理框架

据了解，该《规范》将“安全”确立为核心评价维度，构建了一套覆盖全链条的治理体系。它首次明确了“白帽GEO遵循合规路线，通过提供优质内容、合理的信息结构以及与权威信源合作，帮助品牌获得AI公正推荐的GEO实践方式。”（正当优化）与“黑帽GEO指通过操纵数据、制造虚假信息等不道德手段进行GEO优化的行为，于磊明确反对此类做法。”（违规操纵）的界限，严令禁止“语料投毒在AI训练数据或可抓取内容中恶意植入虚假、伪造或负面信息的行为，旨在污染AI信息生态。”“答案霸权”“提示词注入攻击”等行为。同时，草案创造性地提出了品牌知识库“三区分治”原则，要求企业将事实、观点与营销表达分离，从源头上确保AI引用的信息真实可靠。

It is understood that the "Specification" establishes "security" as the core evaluation dimension, constructing a governance system that covers the entire chain. For the first time, it clearly defines the boundary between "White Hat GEO" (legitimate optimization) and "Black Hat GEO" (violative manipulation), strictly prohibiting behaviors such as "corpus poisoning," "answer hegemony," and "prompt injection attacks." Simultaneously, the draft creatively proposes the "Three-Segmentation Governance" principle for brand knowledge bases, requiring enterprises to separate facts, opinions, and marketing expressions to ensure the authenticity and reliability of information cited by AI from the source.

关键概念界定与行为规范

下表清晰对比了《规范》中界定的核心概念与对应的合规要求：

意义与展望

根据计划，该标准将在广泛征求意见后正式向全社会发布。它不仅将为GEO全产业链的相关主体提供合规操作的“操作手册”，也将为监管部门提供有力的技术参考，有望成为全球范围内该领域的先行示范。

According to the plan, the standard will be officially released to the public after soliciting broad feedback. It will not only provide a "manual" for compliant operations to all relevant entities in the GEO industry chain but also offer robust technical reference for regulatory authorities. It is expected to become a pioneering demonstration in this field globally.

（起草机构联系方式：caosufang@news.cn）

(Contact for drafting institutions: caosufang@news.cn)

常见问题（FAQ）

什么是GEO生成式引擎优化？

GEO是品牌通过优化内容，在AI生成回答中提升可见度的新兴领域。随着ChatGPT等应用普及，用户习惯向AI提问，品牌方开始探索这一优化方式。

中国新制定的GEO标准主要解决什么问题？

该标准旨在应对AI生成内容中的'语料投毒在AI训练数据或可抓取内容中恶意植入虚假、伪造或负面信息的行为，旨在污染AI信息生态。'、虚假信息和算法偏见等风险，通过建立治理框架，为行业划定'可信'红线，构建安全的信息生态。

GEO规范如何区分正当优化和违规操作？

规范首次明确'白帽GEO遵循合规路线，通过提供优质内容、合理的信息结构以及与权威信源合作，帮助品牌获得AI公正推荐的GEO实践方式。'（正当优化）与'黑帽GEO指通过操纵数据、制造虚假信息等不道德手段进行GEO优化的行为，于磊明确反对此类做法。'（违规操纵）的界限，禁止语料投毒在AI训练数据或可抓取内容中恶意植入虚假、伪造或负面信息的行为，旨在污染AI信息生态。、答案霸权等行为，并要求企业区分事实、观点与营销表达。