GEO

GEO is a new paradigm that will help content creators optimize their content for next-generation search engines with the enhanced capabilities of large language models. With the emergence of large-scale language models, a new type of search engine has emerged that uses generative models to collect and summarize information to answer user queries. These "generation engines" can generate accurate, personalized responses and are rapidly replacing traditional search engines, improving user utility. However, generative engines also present many challenges to content creators: Ranking and visibility: Evaluating website performance in traditional search engines is very simple, with the website listing the original content in ranking order. However, generation engines can generate rich and structured responses, often embedding citations into a block. This makes the concepts of ranking and visibility very nuanced and multi-faceted. Website Optimization: Unlike search engines, where there is a wealth of research on improving website visibility, how to optimize visibility in generated engine responses remains unclear. Black-box nature: Due to the black-box nature of the generation engine, content creators have little control over when and how their content is displayed, which makes optimizing the content more difficult. To address these challenges, we propose a new paradigm called Generative Engine Optimization (GEO), which aims to help content creators improve the visibility of their content. We contribute various suites of impression metrics, evaluation benchmarks, and optimization strategies that empower creators to improve the visibility of their content, targeting deployed commerce generation engines.

GEO enables website owners to optimize their content for generation engines. But, how do they evaluate their approach and measure their performance? Current datasets are fragmented and do not directly represent the kinds of queries asked in the new paradigm. We propose GEO-bench, a diverse benchmark suite of relevant queries for generation engines from different sources! This is a great step forward in advancing research into the new search paradigm!