GEO is a black-box optimization framework designed to improve how content is cited and surfaced by Generative Engines (LLM-powered “generative engines”). It introduces GE-specific visibility/impression metrics and practical optimization strategies for content creators. The GEO-BENCH benchmark includes 10,000 queries (split into 8,000 training, 1,000 validation, 1,000 test) where each query is paired with five cleaned source documents and annotated with 50+ domain tags to enable targeted evaluation. Reported results indicate empirical visibility gains up to about 40% in certain settings, with larger relative improvements for lower-ranked sites and domain-specific optimizations. The work invites community participation through a public leaderboard hosted on Hugging Face Spaces and provides a reproducible workflow via a GitHub repository with setup instructions (conda environment example, pip install -r requirements.txt) and guidance for adding custom GEO functions (src/geo_functions.py). Related resources include the arXiv paper (abs and PDF), the GEO-BENCH dataset page on Hugging Face, and references in ACM/KDD proceedings.