Skills / GEO research
OPEN-SOURCE GEO SKILL

Understand how AI
finds your brand.

A GEO skill for Codex and Claude Code. Turn one customer question into a traceable report of search queries, brand mentions, competitor evidence and citations.

Install the skill View source

MIT license · Node.js 22.9+ · Local reports · No key needed for imports

RESEARCH WORKFLOW
“best image generator”
01Collect a real search response
02Inspect visible queries & sources
03Match brand mentions to evidence
Portable report + raw evidence
Try the single-query tool

From a question to evidence you can inspect

Query fan-out

Read the visible query expansion and the search call each query belongs to.

Brand visibility

Match names and aliases in the answer. Separate mentions from brand-owned citations.

Sources & citations

Filter source URLs, compare domains and inspect the actual citation annotations.

Portable evidence

Export HTML, Markdown, normalized JSON and the original response object.

Get started

Install once. Research in your agent.

Install the standalone skill and its bundled runtime with the Skills CLI. Start a new agent session, then invoke $ai-search-lens with a question and your brand.

For Claude Code, use npx skills add PRCrecluse/ai-search-lens-skill --agent claude-code --global to install globally. Omit --global for a project installation.

Terminal · Node.js 22.9+
npx skills add PRCrecluse/ai-search-lens-skill --skill ai-search-lens --agent codex --global

A practical first request

Use $ai-search-lens to research “best image generator”. Track Freebeat, Neural Frames and Kaiber. Save the visible queries, citations and brand evidence as an HTML report.

For a no-cost preview, run npm run report -- --demo --out reports/demo. For live research, configure your API connection locally; do not put a key in prompts or commit it to Git.

What the report can tell you

  • Whether configured brand names appear in this answer.
  • Which public search actions and sources the provider returned.
  • Which source URLs were cited in the final response.
  • What to investigate next, with evidence attached.

This is a snapshot of an API response. It is not hidden reasoning, a personalized ChatGPT session, or a guarantee of future visibility.

GEO skill FAQs

What does this GEO skill do?

It collects or imports a Responses API answer, analyzes visible search queries and brand mentions, traces source URLs and final citations, and exports a portable research report.

Does it work with Codex and Claude Code?

Yes. Install with npx skills add and choose Codex or Claude Code. The standalone skill includes the full Node.js runtime and downloadable release archives.

Do I need an API key?

A key is required for live collection. Demo reports and imports work without one. The CLI supports the OpenAI Responses API and an explicitly configured compatible endpoint. Model and tool availability depend on your provider.

Does it reveal the model’s chain of thought?

No. It records only publicly returned search actions, sources and citations. Those observations do not reveal hidden reasoning or prove why the model chose a brand.

Is one report a visibility score?

No. A single-query report is one answer snapshot. It does not estimate total search volume, a population-level share of voice, or guaranteed ranking improvements.