Skip to content

Latest commit

 

History

History
46 lines (32 loc) · 1.95 KB

File metadata and controls

46 lines (32 loc) · 1.95 KB

OpenCiteX documentation

OpenCiteX is a self-hosted GEO (Generative Engine Optimization) radar. It asks AI search engines the kinds of questions a buyer would ask, then scores whether they mention your brand and cite your domain.

Keys stay on your server (BYOK). One admin password. No SaaS markup.

You want to… Read
Run it on a laptop Getting started
Set env vars and secrets Configuration
Deploy to Vercel + Supabase Deployment
Use Brands, Prompts, scans, and the dashboard Usage
Understand visibility, citation, intercept, rank Scoring
See models, pacing, and the analysis model Engines
Recover a password or fix a failed migrate FAQ

What it measures

Each scan sends your probes to the engines you enable. OpenCiteX stores every answer, then computes:

  • AI visibility — share of unprompted answers that mention you
  • Citation rate — share of those answers that link your official domain
  • Interception — category/scenario probes that name a competitor instead of you
  • Average rank — how early your domain appears in the citation list

Dashboard rates exclude probes that already name your brand, so repeating your name does not inflate the score.

Engines

UI name Model (scan) Search
Perplexity sonar Built in
OpenAI gpt-4o Web search tool
Gemini gemini-3.6-flash Google Search grounding
DeepSeek deepseek-chat Provider search (when available)
Qwen qwen-plus Provider search (when available)

You only pay the providers. OpenCiteX never sees your keys in the browser after save.

Status

MVP: one operator, many brands, no SSO or team roles. Fine for a laptop, a small VPS, or a Vercel hobby project.

License

MIT