AI Visibility Audit

---
name: ai-visibility-audit
description: Audit how a DTC brand appears (or fails to appear) in AI answers using a buyer-question checklist. Use when the user wants a GEO visibility gap analysis without running a monitoring bot.
---

# AI Visibility Audit

## Goal

Produce a structured audit of whether—and how—a brand is likely to be mentioned or cited when buyers ask ChatGPT, Perplexity, Gemini, or Claude category and comparison questions. Output actionable gaps (pages, claims, sources) the team can fix this week.

This skill does **not** call AI search APIs or scrape engines. It uses the operator’s own spot-checks plus on-site evidence.

## Inputs

Ask the user for (or infer from the repo / brand brief):

| Input | Required | Notes |
|-------|----------|-------|
| Brand name + primary domain | yes | e.g. Acme · acme.com |
| Category / niche | yes | e.g. clean protein for women |
| 3–8 buyer questions | yes | Exact phrasings shoppers ask |
| Competitor list (2–5) | recommended | Names + domains |
| Markets / language | recommended | Default: US English |
| Spot-check notes | optional | Paste what ChatGPT/Perplexity said if they already checked |

If buyer questions are missing, propose a starter list (best X, alternatives to Y, brand vs brand, “is Z worth it”, for skin type / use case).

## Steps

1. **Normalize the brand brief**  
   - Canonical name, aliases, product lines, distinct claims (certs, materials, results).  
   - Note what the homepage and top PDPs currently emphasize.

2. **Build the question set**  
   - Split into: category “best of”, problem/solution, comparison, branded, objection.  
   - Mark 3 *must-win* questions.

3. **On-site citation readiness scan** (read pages or ask user for URLs)  
   For each must-win question, check whether the site has:
   - A clear direct answer in the first screenful  
   - Evidence (specs, tests, quotes, comparisons)  
   - Stable, linkable URL (not only Instagram)  
   - FAQ / comparison / glossary coverage  
   Score each: Strong / Partial / Missing.

4. **Competitor narrative sketch**  
   - From public knowledge + user spot-checks: how competitors are typically described (hero claim, proof, price tier).  
   - Flag claims competitors own that the brand never states on-site.

5. **Spot-check protocol** (for the human or agent browser, if available)  
   - Run each must-win question once in 1–2 engines.  
   - Record: mentioned? position in list? cited URL? recommended vs named only?  
   - Do **not** invent engine results—if unchecked, mark `unchecked`.

6. **Synthesize gaps**  
   - Map each weak/missing score to a concrete content or structural fix.  
   - Prioritize fixes that unblock must-win questions.

## Output format

Return Markdown:

```markdown
# AI Visibility Audit — {Brand}

## Brief
- Domain:
- Category:
- Must-win questions:

## Question matrix
| Question | Type | On-site readiness | Spot-check (engine) | Gap |
|----------|------|-------------------|---------------------|-----|

## Top 5 gaps (prioritized)
1. ...

## 7-day fix plan
- Day 1–2: ...
- Day 3–5: ...
- Day 6–7: ...

## Suggested next skills
- citation-worthy-page / faq-comparison-geo / ...
```

## Guardrails

- Never claim guaranteed rankings or citations.  
- Never fabricate ChatGPT/Perplexity outputs; use `unchecked` or user-provided paste.  
- Prefer fixes the brand controls (site structure, evidence, FAQs) over “get more backlinks” fluff.  
- Keep recommendations DTC-practical (Shopify-friendly page types).