> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xpoz.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# GEO Visibility Check

> One-shot audit of whether AI engines recommend your brand: per-prompt verdicts, who wins instead, and which sources the answers come from.

The GEO Visibility Check skill answers the first question of any GEO (Generative Engine Optimization) effort: for the buyer questions that matter, does the brand appear in Claude, ChatGPT, and Gemini answers, and if not, who does and through which sources? It assembles a small panel of buyer prompts, runs each through the engines with full citation tracing, and delivers a per-prompt visibility verdict. One-shot by design: it is the measurement pass that tells you whether a full GEO program is worth running.

## Example Prompts

* "Do AI assistants recommend \[brand]?"
* "GEO audit for \[brand]: 8 prompts, all three engines"
* "Who wins \[category] in AI answers, and why?"
* "Which sources do AI engines assemble \[category] answers from?"

## Audit Methodology

<Steps>
  <Step title="Build the prompt panel">
    3-8 questions buyers actually ask assistants, derived from the product's category, personas, and competitor set (community mining via Xpoz helps surface real phrasings).
  </Step>

  <Step title="Trace every prompt">
    Each prompt runs through Claude, ChatGPT, and Gemini with live web search on, capturing answers, the engines' search queries, and cited URLs (the ai-answer-trace skill provides the tracing).
  </Step>

  <Step title="Judge visibility per prompt">
    Present (recommended, listed, mentioned, or absent), cited (brand-owned URL in the citations), who wins instead and via which URLs, and stability across samples.
  </Step>

  <Step title="Report">
    Per-prompt verdict table, cited-domain rankings with surface types, brand citation paths, lost prompts diagnosed, and recommended next steps.
  </Step>
</Steps>

## Requirements

Engine tracing needs AI engine API keys (any of Anthropic, OpenAI, Gemini; one is enough to start). The demand-research side needs only Xpoz. Engine answers are nondeterministic, so verdicts use 2+ samples per engine, and keeping the trace files makes a re-run next month a real before/after.

## Xpoz Tools Used

| Tool                        | Purpose                                        |
| --------------------------- | ---------------------------------------------- |
| `getRedditPostsByKeywords`  | Mine real buyer phrasings for the prompt panel |
| `getTwitterPostsByKeywords` | Same, on X                                     |

## Get Started

Get a free access key at [xpoz.ai/get-token](https://xpoz.ai/get-token). See the [installation guide](/skills/overview#setup) to add this skill to your agent. For the weekly program this check previews (tracked prompt panel, dated snapshots, gap analysis, content production), use the open-source [geo-seo-agent](https://github.com/XPOZpublic/geo-seo-agent).