[论文解读] AgenticGEO: A Self-Evolving Agentic System for Generative Engine Optimization

Jiaqi Yuan, Jialu Wang+4 more|arXiv (Cornell University)|Mar 2, 2026

Information Retrieval and Search Behavior被引用 0

一句话总结

AgenticGEO 将 GEO 形式化为一个内容条件优化问题，并使用共进化的 MAP-Elites 策略档案加上轻量级评判来适应黑箱生成引擎，在减少引擎反馈的情况下取得了最先进的结果。

ABSTRACT

Generative search engines represent a transition from traditional ranking-based retrieval to Large Language Model (LLM)-based synthesis, transforming optimization goals from ranking prominence towards content inclusion. Generative Engine Optimization (GEO), specifically, aims to maximize visibility and attribution in black-box summarized outputs by strategically manipulating source content. However, existing methods rely on static heuristics, single-prompt optimization, or engine preference rule distillation that is prone to overfitting. They cannot flexibly adapt to diverse content or the changing behaviors of generative engines. Moreover, effectively optimizing these strategies requires an impractical amount of interaction feedback from the engines. To address these challenges, we propose AgenticGEO, a self-evolving agentic framework formulating optimization as a content-conditioned control problem, which enhances intrinsic content quality to robustly adapt to the unpredictable behaviors of black-box engines. Unlike fixed-strategy methods, AgenticGEO employs a MAP-Elites archive to evolve diverse, compositional strategies. To mitigate interaction costs, we introduce a Co-Evolving Critic, a lightweight surrogate that approximates engine feedback for content-specific strategy selection and refinement, efficiently guiding both evolutionary search and inference-time planning. Through extensive in-domain and cross-domain experiments on two representative engines, AgenticGEO achieves state-of-the-art performance and demonstrates robust transferability, outperforming 14 baselines across 3 datasets. Our code and model are available at: https://github.com/AIcling/agentic_geo.

研究动机与目标

• 在非平稳的黑箱 GE 行为中，提升生成引擎输出（GEO）的可见性与归因性的动机。
• 开发一个自我进化的框架，能够灵活地将改写策略适配到多样化内容。
• 在保持跨领域鲁棒优化的同时，降低对成本高昂的引擎反馈的依赖。

提出的方法

• 将 GEO 形式化为一个内容条件的优化问题。
• 维护一个高质量-多样性 MAP-Elites 改写策略档案，以覆盖多样内容。
• 引入一个共进化的轻量级评判来引导策略选择与推理规划。
• 离线评判对齐以通过初始档案中的监督目标进行引导。
• 在线对档案与评判进行共进化，在有限的 GE 反馈下工作。
• 在推理阶段进行代理式多轮改写，由评判引导的规划驱动。

Figure 1. Characterization of the GEO result on GEO-Bench instances. $y$ -axis reports maximum performance among 9 rewriting strategies, and $x$ -axis reports performance variance among strategies. (i) Optimization success varies greatly by strategy and content. (ii) Existing strategies fail to opti

实验结果

研究问题

• RQ1 AgenticGEO 相对于最先进的 GEO 基线在不同引擎与领域的表现如何？
• RQ2 该方法能否迁移到未见领域同时保持性能？
• RQ3 每个共进化组件（档案、评判、 Evolver）对性能的影响是什么？
• RQ4 优化在提升可见性与归因性的同时是否保留语义含义？

主要发现

方法	GEO-Bench 词	GEO-Bench 词性	GEO-Bench 总体	Llama-3.3-70B-Instruct 词	Llama-3.3-70B-Instruct 词性	Llama-3.3-70B-Instruct 总体
无优化	20.05	20.26	20.21	19.19	19.33	19.20
关键词堆叠	20.73	20.86	20.69	19.99	20.16	20.02
独特词汇	17.59	17.94	17.78	16.78	16.66	16.56
易于理解	20.10	20.19	20.05	18.72	18.93	18.85
权威性	20.41	20.93	20.60	19.41	19.48	19.47
技术词汇	21.22	20.97	21.23	19.55	19.59	19.50
流畅性优化	20.66	20.85	20.70

• AgenticGEO 在两个具代表性的生成引擎上实现了最先进的性能，相对于基线的平均增益达到 46.4%。
• 仅使用 41.2% 的 GE 反馈就保持了 98.1% 的性能，展示了监督需求的下降。
• 在3个数据集的内域与跨域设置中，优于14个基线方法。
• 不断进化的评判能够有效替代昂贵的引擎反馈，并在引擎架构与规模上具有泛化能力。

Figure 2. GEO v.s. AgenticGEO. Static GEO methods apply fixed rewriting heuristics, whereas AgenticGEO maintains an evolving strategy archive and a critic to adaptively retrieve high-scoring strategies for iterative rewriting.