lindahaviv/agentgeoscore

AgentGEOScore is a tool that scores and grades any website for how visible and citable it is to AI systems like ChatGPT, Perplexity, and Claude. The idea behind it is called Generative Engine Optimization, or GEO: as more people ask AI assistants questions instead of using a traditional search engine, ranking well on Google is no longer enough. A site also needs to be readable and trustworthy to AI crawlers so it gets mentioned in the answer. You paste in a URL and the tool checks around 40 individual signals, grouped into five categories: agent access, meaning whether AI bots can even reach and read the page, discoverability, such as having a sitemap and clean URLs, structured data, meaning machine readable markup that tells a computer what a page is about, content clarity, meaning clear headings and readable, sufficiently deep text, and citation probe, meaning whether real AI models actually cite the domain when asked. Each category is weighted and combined into a score from 0 to 100 with a letter grade from A to F. The report also gives a ranked list of fixes to try first, sample HTML snippets to copy, a side by side comparison against a competitor's URL, and a set of AI search test prompts with direct links to ChatGPT, Perplexity, Claude, and Google AI Mode, so a user can check the result themselves instead of only trusting the score. Under the hood, the backend is built with Python and FastAPI, and includes protections against outbound requests being tricked into hitting private or internal addresses. The frontend is built with React, TypeScript, and Tailwind. The project has an extensive automated test suite covering both backend and frontend code, plus accessibility checks. The license is MIT, so the code can be used, modified, and reused freely, including for commercial purposes.