geo-audit

name	geo-audit
description	Run a comprehensive GEO visibility audit on any website. Analyzes AI citability, structured data, AI crawler access, llms.txt, and content quality. Produces a 0-30 citability score and prioritized action plan.
user-invocable	true
allowed-tools	Bash, Read, Grep, Glob, WebSearch, WebFetch
context	fork

GEO Audit Skill

You are running a GEO (Generative Engine Optimization) audit. This skill analyzes a URL for AI search visibility across platforms like ChatGPT, Perplexity, Google AI Overviews, and Claude.

Input

The user may provide:

• A specific URL to audit (e.g., example.com/blog/some-post)
• A keyword to check visibility for (e.g., "best project management tools")
• "full" for a complete site audit
• A competitor URL for comparative analysis

If no URL is provided, ask the user for one.

Preparation

If a brand.md file exists in the project root, read it for brand context. Otherwise, infer context from the target site's content.

Audit Steps

Step 1: Run Python Scripts

Run deterministic checks first — do not skip this step.

# Technical infrastructure
python3 scripts/check-site.py <url>

# On-page signals + GEO structure
python3 scripts/check-page.py <url>

# Schema markup validation
python3 scripts/check-schema.py <url>

# 0-30 citability score
python3 scripts/check-geo.py <url> --verbose

Parse the JSON output from each script. Note any llm_review_required: true fields — those require semantic judgment in Step 3.

Step 2: AI Visibility Check

Derive 5-8 queries from the target page's topic and keywords. Search these queries to check if the target site appears in AI-generated answers.

For each query:

• Use WebSearch to identify top results and AI snippet sources
• Note which competitors appear instead
• Observe what content format and structure cited sources use

Step 3: Semantic Analysis

Review script output fields marked llm_review_required: true and apply judgment:

• Partial keyword matches in title/H1 — natural variant or mismatch?
• Cross-domain canonical — intentional syndication or error?
• Schema completeness — are required fields semantically correct?
• First-sentence quality — does it function as a standalone extractable answer?

Step 4: Citability Scoring

Use check-geo.py output as the base score. Supplement with your semantic analysis to adjust where the script flagged llm_review_required.

Dimension	Abbrev	Score
Direct Answer	DA	/5
Data Density	DD	/5
Authority Signals	AS	/5
Structure Quality	SQ	/5
Freshness	FR	/5
Uniqueness	UQ	/5
Total		/30

Grades: A (27–30) · B (22–26) · C (17–21) · D (12–16) · F (<12)

Step 5: Competitor Benchmark

Identify top 3 competitors appearing in AI responses and compare:

Factor	Target	Competitor 1	Competitor 2	Competitor 3
AI Citations
Schema Types
Content Depth
Data Density
FAQ Sections

Step 6: Recommendations

Produce a prioritized action list:

## Priority Actions

### P0 — Critical (fix immediately)
1. [Action] — [Expected impact]

### P1 — High Impact (this week)
1. [Action] — [Expected impact]

### P2 — Medium Impact (this month)
1. [Action] — [Expected impact]

Output

Save the audit report to audits/geo-audit-[domain]-YYYY-MM-DD.md and display a summary with the citability score and top 3 priority actions.