AgenticGEO: A Self-Evolving Agentic System for Generative Engine Optimization
arXiv:2603.20213v1 Announce Type: new 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 qua
arXiv:2603.20213v1 Announce Type: new 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.
Executive Summary
This article proposes AgenticGEO, a self-evolving agentic system for Generative Engine Optimization (GEO), which aims to maximize visibility and attribution in black-box summarized outputs by strategically manipulating source content. Unlike existing methods, AgenticGEO employs a MAP-Elites archive to evolve diverse, compositional strategies and introduces a Co-Evolving Critic to mitigate interaction costs. Through extensive experiments, AgenticGEO achieves state-of-the-art performance and demonstrates robust transferability, outperforming 14 baselines across 3 datasets. The proposed system has the potential to revolutionize GEO by enabling flexible adaptation to diverse content and changing engine behaviors, reducing the need for impractical amounts of interaction feedback.
Key Points
- ▸ AgenticGEO is a self-evolving agentic system for Generative Engine Optimization (GEO).
- ▸ AgenticGEO employs a MAP-Elites archive to evolve diverse, compositional strategies.
- ▸ The Co-Evolving Critic is introduced to mitigate interaction costs and efficiently guide evolutionary search and inference-time planning.
Merits
Strength in Adaptability
AgenticGEO's ability to flexibly adapt to diverse content and changing engine behaviors is a significant strength, enabling it to outperform existing methods in various experiments.
Efficient Interaction
The Co-Evolving Critic effectively mitigates interaction costs, reducing the need for impractical amounts of feedback from the engines.
Robust Transferability
AgenticGEO demonstrates robust transferability across different datasets and engines, showcasing its potential for real-world applications.
Demerits
Complexity
The proposed system's complexity may be a limitation, particularly for implementation and maintenance, especially for those without extensive experience in agentic systems and Generative Engine Optimization.
Scalability
AgenticGEO may not scale well to very large datasets or complex engine behaviors, potentially limiting its applicability in real-world scenarios.
Expert Commentary
AgenticGEO is a pioneering work in the field of Generative Engine Optimization (GEO), showcasing the potential of self-evolving agentic systems to adapt to diverse content and changing engine behaviors. The proposed system's ability to efficiently guide evolutionary search and inference-time planning, while mitigating interaction costs, is a significant strength. However, its complexity and potential scalability limitations may be a concern for real-world applications. Nevertheless, AgenticGEO's robust transferability and state-of-the-art performance make it an exciting area of research, with potential implications for explainability in generative models, transfer learning in AI systems, and policy-making.
Recommendations
- ✓ Future research should focus on addressing the complexity and scalability limitations of AgenticGEO, potentially by exploring more efficient optimization algorithms or developing more robust and adaptable Co-Evolving Critics.
- ✓ The proposed system's potential implications for explainability in generative models and transfer learning in AI systems should be further explored, particularly in real-world applications and policy-making contexts.
Sources
Original: arXiv - cs.AI
