TECHNICAL_OVERVIEW

Submit a URL and GEO Audit fetches the page content via Cheerio and Turndown (HTML to Markdown, capped at 30K characters), then scores it across three audit tiers. The free basic audit evaluates three core metrics — Fact Density, Entity Salience, and Extractability — and returns an overall GEO score with AI citation simulation, findings, and recommendations. The advanced audit (1 credit) expands to eight metrics, adding Source Citation, Schema Markup, Direct Answer Formatting, Multimedia Optimization, and Conversational Flow, with an executive summary, citation examples, competitor gap analysis, and platform fit assessment. The commerce audit (2 credits) scores against six weighted protocols — Universal Baseline (40%), Google UCP (20%), OpenAI ACP (15%), Amazon Buy for Me (10%), Perplexity (10%), and Microsoft Copilot (5%) — each running 7–9 binary checks to produce a readiness verdict: discovery-ready at 50+, transaction-ready when any non-Universal protocol also scores 60+.

All three audit types stream results via SSE. The frontend and backend are fully decoupled: React 19 on Vite 7 communicates with an Express 4 API on Node 22, with VITE_* vars baked into the browser bundle at build time and all API keys restricted to the server layer. LLM model selection (OpenAI, Gemini, Anthropic) and prompt versions are managed through a database-backed admin UI with version history, so model and prompt changes deploy without code changes. JWT authentication enforces a role hierarchy (user/admin/super_admin). Stripe handles credit purchases. Results render as Recharts visualisations with PDF export. Built with React 19, TypeScript, Vite 7, Tailwind CSS 4, Express 4, PostgreSQL, Zod, Docker, Nginx, and deployed on Railway.