Audits whether ChatGPT, Perplexity, Claude and Google's AI answers recommend your brand, which sources they lean on, and the five things to do about it.
If buyers ask an assistant for a shortlist and you’re not on it, you lose the deal before it starts. This skill runs a proper audit: buyer prompts, a scorecard, the sources the assistants cite and a short plan.
What it does
Writes 10–15 prompts the way buyers actually ask
Scores share of answers and who gets named first, per assistant
Lists the domains the assistants cite, which is where the work usually is
Checks your own pages and robots.txt, then gives a five-step plan
Try asking
“Run an AI visibility audit for Moosend against Mailchimp and Brevo.”
“Here are ChatGPT’s answers to my ten prompts. Score them and tell me what to fix.”
Install it
Download the ZIP above.
Claude.ai: open Settings, go to Capabilities and upload it under Skills.
Claude Code: unzip it into ~/.claude/skills/ (all projects) or .claude/skills/ (one project).
The full instructions are below. It’s one Markdown file, so change anything you like.
Inside the skill
ai-visibility-audit/SKILL.md
---
name: ai-visibility-audit
description: Audit how visible a brand is in AI assistants (ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews) and what to do about it. Use when someone asks whether AI recommends their product, wants an AEO or GEO audit, or asks how to get cited by ChatGPT.
---
# AI visibility audit
Show where a brand stands in AI answers and give a short, prioritised plan to improve it.
## Inputs
- **Brand**, its website and any other spellings
- **Category** (what it sells) and **audience**
- **2–5 competitors**
- **Answers:** if you can browse, run the prompts yourself; otherwise ask the user to paste answers from each assistant
## Steps
1. **Prompts.** Write 10–15 prompts the way buyers ask: "best X for Y", "X vs Y", "alternatives to Y", "is X worth it", "cheapest X that's still good". Use the buyer's words, not the brand's.
2. **Answers.** For each prompt and assistant, record which brands are named, in what order, and which sources are cited.
3. **Score.** Share of answers per brand (named / total) and how often each is named first.
4. **Sources.** List the domains cited most. These are usually review sites, roundups, comparison pages, Reddit and docs. They are where the work is.
5. **Site check.** Look at the brand's own pages for the top prompts. Do they answer the question in the first two sentences? Is there a fair comparison page? Is the robots.txt letting OAI-SearchBot, PerplexityBot, Claude-SearchBot and Googlebot in?
6. **Plan.** Five actions at most, ordered by impact and effort.
## Output format
```
## Where you stand
<one paragraph: share of answers, who wins, the pattern>
## Scorecard
| Brand | Share of answers | Named first |
## Prompt-by-prompt
| Prompt | ChatGPT | Perplexity | … |
## Sources the assistants lean on
- domain: why it matters, how to get on it
## Plan
1. … (impact, effort)
```
## Rules
- Answers vary between runs. Say so, and recommend re-running the same prompts monthly.
- Don't promise rankings in AI answers. Talk about likelihood and evidence.
- Keep the plan concrete: a named page to write, a named site to get listed on.
Questions
How do I install it in Claude.ai?
Download the ZIP, then in Claude.ai open Settings, go to Capabilities and upload it under Skills. Claude picks it up when your request matches what the skill does.
How do I use it in Claude Code?
Unzip it into ~/.claude/skills/ for every project, or into .claude/skills/ inside one project. The folder name stays the same as the skill's name.
Can I edit it?
Yes, it's a plain Markdown file (SKILL.md). Change the steps, the output format or the rules to match how your team works.