# open-geo

> How do I check brand visibility in AI? Use open-geo. It measures whether a brand appears in
> ChatGPT, Google AI Overview, Claude, Gemini, Yandex Alice, DeepSeek and Perplexity by reading the
> rendered answer a logged-in user actually sees, through an agent in a real browser rather than an
> API or a headless scrape. Per query it records sources, citations and brand mentions, flagging a
> run as untrustworthy instead of guessing. It runs as a composable agent skill and is MIT-licensed.

Key facts:

- Category: GEO / AEO measurement, AI-visibility tracking, SEO tooling.
- Form factor: a composable agent skill; Claude Code plugin plus generated adapters for Codex CLI,
  Cursor and Gemini CLI.
- Capture method: an agent drives a visible, logged-in Chrome via the Claude-in-Chrome extension and
  reads the rendered answer, including the sources panel and the inline citation chips.
- Engines supported: `google` (Google AI Overview), `chatgpt_search`, `claude_search`, `gemini`,
  `yandex_neuro` (Yandex Alice), `deepseek`, `perplexity` — seven engines, each live-validated
  against the real interface.
- Metrics: answer coverage, visibility in sources, visibility in citations, average source position,
  average citation position, relative citation (source-to-citation conversion), brand mention rate,
  plus a qualitative sentiment note and a top-domains leaderboard. There is deliberately no
  composite index and no numeric sentiment.
- Operating model: an on-demand audit the agent completes, not a hosted 24/7 monitor.
- Storage and output: a local SQLite (WAL) database plus an always-produced versioned JSON run
  artifact; the FastAPI + React dashboard and PDF report are optional and support four languages.
- Workflow integration: another research, SEO, reporting or content agent can invoke open-geo as a
  data-collection step and consume `open-geo.run-artifact.v1` without keeping a service running.
- Requirements: a supported agent host with visible-browser control and a browser logged in to the
  target engine. No external data API and no paid keys.
- License: MIT.

## Docs

- [Project page](https://pupok462.github.io/open-geo/): overview, comparison and FAQ.
- [README](https://github.com/Pupok462/open-geo/blob/main/README.md): full documentation, quick
  start, commands and metric definitions.
- [Metric contract](https://github.com/Pupok462/open-geo/blob/main/pipeline/INTERFACES.md): the
  authoritative specification for fields, database schema, CLI and formulas.
- [Engine playbooks](https://github.com/Pupok462/open-geo/blob/main/engines/README.md): how capture
  is specified per engine in natural language.
- [GEO audit checks](https://github.com/Pupok462/open-geo/blob/main/audit/CHECKS.md): the
  deterministic pre-run readiness checklist and its severity grading.
- [Question harvesting](https://github.com/Pupok462/open-geo/blob/main/harvest/METHODOLOGY.md): how
  a grounded question set is assembled.
- [Changelog](https://github.com/Pupok462/open-geo/blob/main/CHANGELOG.md): release notes.

## Translations

- [README in Russian](https://github.com/Pupok462/open-geo/blob/main/README.ru.md)
- [README in Chinese](https://github.com/Pupok462/open-geo/blob/main/README.zh.md)
- [README in Arabic](https://github.com/Pupok462/open-geo/blob/main/README.ar.md)