Agent2UCB Optimizes Content for Generative AI Search Engines.

The emergence of large language model-driven search engines, such as Google AI Overviews and Perplexity, has created a new field called Generative Engine Optimization (GEO). This practice focuses on refining content to maximize its chances of being cited or summarized by these generative AI systems. This research introduces Agent2UCB, an agentic system specifically designed for GEO. Agent2UCB operates autonomously, continuously improving content visibility through a feedback-driven optimization process. For each piece of content, the system assesses nine distinct GEO strategies. It then employs a bandit-based Agent2UCB policy, which combines LLM priors with real-time reward signals, to efficiently identify and select the most effective optimization method. Beyond GEO, the system also incorporates a lightweight, text-only SEO readiness evaluation. This monitors potential side effects by checking for readability, topical coverage, and EEAT-style credibility, ensuring that GEO gains do not compromise traditional SEO quality. Experiments on the GEO-Bench dataset demonstrate consistent visibility improvements while maintaining SEO standards, offering a practical tool for content creators and marketers navigating the evolving search landscape.