Difficulty: Intermediate Time to complete: 10 minutes per audit

Use Claude Code to audit any company's AI search visibility in under 10 minutes. Check schema markup, AI citations, content structure, and more.

---

Why This Matters

Here is a stat that should make every marketer uncomfortable: only about 12% of SaaS company websites have proper schema markup. Only 8% have structured FAQ sections. That means the vast majority of businesses are invisible to AI search engines like ChatGPT, Perplexity, and Google's AI Overviews.

Traditional SEO audits check backlinks and keyword density. An AI visibility audit checks whether AI systems can actually find, understand, and cite your content. These are different things, and most companies are only doing the first one.

This playbook shows you how to build an audit skill in Claude Code that checks all five critical areas in about 10 minutes. You can run it on your own site, your client's site, or a prospect's site to open a sales conversation.

What You Will Build

A Claude Code skill that fetches a website, analyzes its AI-readiness across five checkpoints, and generates a scored report you can share with clients or use internally.

Tools used: Claude Code (terminal), web search, web fetch, skill files

---

Step 1: Set Up the Audit Skill

Create your skill file at skills/ai-visibility-audit.md:

# AI Visibility Audit Skill

## Purpose
Audit any website for AI search visibility. Check whether AI systems
(ChatGPT, Perplexity, Google AI Overviews) can find, parse, and cite
this company's content.

## Audit Process
When asked to audit a website:
1. Fetch the homepage and 2-3 key pages (about, blog, product/service page)
2. Run all 5 checkpoint analyses
3. Score each checkpoint: PASS, PARTIAL, or FAIL
4. Generate a summary report with specific recommendations

## The 5 Audit Checkpoints

### Checkpoint 1: Schema Markup
- Check for JSON-LD structured data on the homepage
- Look for: Organization, Product, FAQ, HowTo, Article schemas
- Check if schema is valid and complete (not just present but empty)
- PASS: Has 2+ schema types with complete data
- PARTIAL: Has schema but incomplete or only 1 type
- FAIL: No schema markup found

### Checkpoint 2: Meta and Open Graph Tags
- Check for proper meta descriptions (not boilerplate)
- Check Open Graph tags (og:title, og:description, og:image)
- Check for canonical URLs
- Check for proper title tag structure
- PASS: All meta tags present and descriptive
- PARTIAL: Some tags missing or generic
- FAIL: Missing key meta tags

### Checkpoint 3: Content Structure
- Check heading hierarchy (H1 > H2 > H3, properly nested)
- Check for FAQ sections with clear question/answer format
- Check for "What is [product/concept]" definitional content
- Check for comparison content ("X vs Y")
- Check content length on key pages (thin content = invisible content)
- PASS: Well-structured content with FAQ and definitional sections
- PARTIAL: Good structure but missing FAQ or definitional content
- FAIL: Poor heading structure, thin content, no FAQ

### Checkpoint 4: AI Citation Test
- Search Perplexity for the company name + primary keyword
- Search ChatGPT (via web) for the same query
- Check if the company appears in AI-generated answers
- Note: This is a manual check, document what you find
- PASS: Cited in 2+ AI engines
- PARTIAL: Cited in 1 AI engine
- FAIL: Not cited anywhere

### Checkpoint 5: Robots.txt and AI Crawler Access
- Fetch /robots.txt
- Check if AI crawlers are blocked (GPTBot, ClaudeBot, Amazonbot, etc.)
- Check for sitemap reference
- Check crawl-delay settings
- PASS: AI crawlers allowed, sitemap present
- PARTIAL: Some crawlers blocked or no sitemap
- FAIL: All AI crawlers blocked

## Report Format
Generate a markdown report with:
- Company name and URL
- Date of audit
- Overall score (out of 5 checkpoints)
- Each checkpoint: status, findings, specific recommendation
- Priority action items (top 3 things to fix first)
- Competitor comparison note (if requested)

Step 2: Update Your CLAUDE.md

Add the audit skill to your CLAUDE.md skill directory:

## Skill Files
- `skills/ai-visibility-audit.md` — AI visibility audit for any website

If you already have a CLAUDE.md from the LinkedIn post playbook, just add this line to your existing skill files section.

Step 3: Run Your First Audit

Start Claude Code in your marketing-system folder and give it a target:

Run an AI visibility audit on example.com

Claude Code will read your skill file and begin the process. It will fetch the site, check each of the five areas, and generate a report.

Here is what happens under the hood for each checkpoint:

Schema check: Claude Code fetches the page source and looks for <script type="application/ld+json"> blocks. It parses the JSON and checks what schema types are present and whether the data fields are actually populated.

Meta tag check: It reads the HTML head section for meta descriptions, OG tags, canonical URLs, and title tags. It flags anything generic like "Welcome to our website" as a problem.

Content structure check: It analyzes the heading hierarchy, looks for FAQ-formatted content, and checks whether key pages have enough substance to be useful to an AI trying to answer questions about the company.

AI citation check: This one requires web search. Claude Code will search Perplexity and ChatGPT-style queries to see if the company shows up. You can also do this manually and feed the results back.

Robots.txt check: It fetches /robots.txt and checks the crawler rules.

Step 4: The AI Citation Test (Manual Prompts)

For Checkpoint 4, you will want to test these specific prompts in Perplexity and ChatGPT. Here are the templates:

Perplexity test prompts:

What is [company name] and what do they do?

What are the best [product category] tools for [use case]?

Compare [company name] vs [competitor name]

ChatGPT test prompts:

Tell me about [company name]. What products do they offer?

What companies should I consider for [product category]?

What do customers say about [company name]?

Run each of these and note:

• Does the company appear at all?
• Is the information accurate and current?
• Are they cited as a source, or is the information pulled from third-party reviews?

Feed these findings back to Claude Code:

Here are the AI citation results for example.com:
- Perplexity: Found in 1 of 3 queries, information was outdated
- ChatGPT: Not found in any query
Update the audit report with these results.

Step 5: Generate the Client-Ready Report

Once all five checkpoints are complete, ask Claude Code to format the final report:

Generate the final audit report for example.com.
Format it as a professional document I can send to a prospect.
Include the overall score, each checkpoint result, and the top 3 priority fixes.

Here is what a sample report structure looks like:

# AI Visibility Audit: Example.com
**Date:** February 27, 2026
**Overall Score:** 2/5

## Summary
Example.com has significant gaps in AI search visibility.
The site lacks structured data, has no FAQ content, and is not
currently cited by any major AI search engine. Three targeted
fixes could dramatically improve visibility within 30 days.

## Checkpoint Results

### 1. Schema Markup: FAIL
No JSON-LD structured data found on any page.
**Fix:** Add Organization, Product, and FAQ schema to homepage
and key product pages. Estimated time: 2-4 hours.

### 2. Meta Tags: PARTIAL
Meta descriptions present but generic. OG tags missing on blog posts.
**Fix:** Write unique meta descriptions for top 20 pages.
Add OG tags to blog template.

### 3. Content Structure: FAIL
No FAQ sections. No definitional content. Blog posts average
300 words with no heading structure.
**Fix:** Add FAQ section to homepage and top 5 product pages.
Expand key blog posts to 1,000+ words with proper H2/H3 structure.

### 4. AI Citations: FAIL
Not cited in Perplexity or ChatGPT for any tested query.
**Fix:** Create "What is [product]" content. Build comparison
pages. Get mentioned in industry roundups and reviews.

### 5. Robots.txt: PASS
AI crawlers are not blocked. Sitemap is present and referenced.
No issues found.

## Priority Actions
1. Add schema markup to homepage (biggest quick win)
2. Create FAQ content on top 5 pages
3. Write one comprehensive "What is [product category]" guide

Using This for Prospecting

This audit is one of the most effective cold outreach tools available right now. Here is why: you are showing a prospect something they have never seen before. Nobody is sending them AI visibility audits. Everyone is sending them SEO audits.

Run the audit before you reach out. Include the key findings in your first message. "I ran a quick AI visibility audit on your site. You are not showing up in Perplexity or ChatGPT for your primary keywords, and I found three specific things that would fix that." That opens doors.

The stats back this up. When only 12% of SaaS companies have schema and 8% have FAQ sections, nearly every prospect you audit will have clear gaps you can point to. That is not a pitch. That is a diagnosis.

Scaling the Audit

Once your skill file is dialed in, you can audit multiple sites in a session:

Run AI visibility audits on these 5 companies:
1. company-a.com
2. company-b.com
3. company-c.com
4. company-d.com
5. company-e.com

Generate a comparison table showing how they stack up against each other.

This is particularly useful for competitive analysis or when pitching to multiple companies in the same industry.

File Structure

marketing-system/
  CLAUDE.md
  skills/
    ai-visibility-audit.md
  reports/
    audit-example-com-2026-02-27.md

Keep your reports in a dedicated folder. They become a portfolio of work you can reference and a dataset you can use to spot patterns across industries.

---

Tools used in this playbook: Claude Code CLI, web fetch, web search, skill files, markdown reporting