> ## Documentation Index
> Fetch the complete documentation index at: https://areyoufoundbyai.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Run a Six-Dimension GEO Audit with the geo-audit Skill
> Run 5 parallel subagents across 6 GEO dimensions, compute thl-open's own composite GEO Score (not the Are you found by AI? Readiness score), and write a prioritised action plan to GEO-AUDIT-REPORT.md.
The `geo-audit` skill is the orchestration layer at the heart of thl-open. Point it at any URL and it runs five specialised subagents in parallel (one per GEO dimension group), each fetching evidence, scoring its dimensions and returning a provenance-tagged result. The skill then computes a composite GEO Score (0 to 100) and writes everything into a single `GEO-AUDIT-REPORT.md` you can act on.
**This composite is not the AI Readiness score in your Are you found by AI? report.** The geo-audit GEO Score is thl-open's own audit: six dimensions, including Brand Authority, combined with the repo's fixed weights. That AI Readiness score is a different measure, composed of five site signals (Citability, E-E-A-T, Technical, Schema and Platform), with off-site presence scored separately as Footprint. Do not compare the two numbers directly or present one as the other.
Like AI Readiness, the geo-audit composite is a **readiness** measure: it reads public signals and infers how citable and recommendable a site is. It does **not** query the AI engines to confirm whether they actually name you right now. For that, use the `found-by-ai` skill or run the free scan at [areyoufoundbyai.com](https://areyoufoundbyai.com).
## Prerequisites
* **Claude Code** with the THL Open skill suite installed:
```bash theme={null}
npx skills add techhorizonlabs/thl-open
```
## How to Run a Full Audit
Invoke the skill from Claude Code with the target URL:
```text theme={null}
/geo audit https://example.com
```
Claude Code begins a three-phase workflow:
The skill fetches the homepage, detects your business type (SaaS, Local Business, E-commerce, Publisher, Agency, or Professional/YMYL), then crawls your sitemap or internal links to build a page list of up to 50 URLs. It records titles, headings, schema markup, word counts, and response codes for every page.
Five specialised subagents run **simultaneously** (not sequentially) so the full audit completes in roughly the time it takes to run the slowest single dimension:
| Subagent | Dimensions measured |
| - | - |
| **AI Visibility Analysis** | AI Citability (25%) + Brand Authority (20%) |
| **Platform Optimization** | Platform Optimization (10%) |
| **Technical GEO Infrastructure** | Technical GEO (15%) |
| **Content E-E-A-T Quality** | Content E-E-A-T (20%) |
| **Schema & Structured Data** | Schema & Structured Data (10%) |
Every subagent returns its dimension score (0 to 100, or a placeholder dash instead of a number when unmeasured), a **provenance tag**, and findings tied to the evidence that justifies them.
The orchestrator collects all subagent results, applies the fixed composite formula, and writes `GEO-AUDIT-REPORT.md` to your working directory.
## The Composite Formula
The overall GEO Score is a weighted average across six dimensions. These are thl-open's published weights for its own audit, and subagents cannot change them:
| Dimension | Weight | What it measures |
| - | - | - |
| AI Citability | 25% | How quotable and extractable your content is for AI systems |
| Brand Authority | 20% | Third-party mentions and entity recognition signals |
| Content E-E-A-T | 20% | Experience, Expertise, Authoritativeness, Trustworthiness |
| Technical GEO | 15% | AI crawler access, `llms.txt`, rendering, and page speed |
| Schema & Structured Data | 10% | Schema.org markup quality and completeness |
| Platform Optimization | 10% | Presence on platforms AI models train on and cite |
```
GEO_Score = (Citability × 0.25) + (Brand × 0.20) + (EEAT × 0.20)
+ (Technical × 0.15) + (Schema × 0.10) + (Platform × 0.10)
```
## Output: GEO-AUDIT-REPORT.md
The skill writes a structured Markdown file containing:
Overall GEO Score (0 to 100) with a rating band: Excellent (90 to 100), Good (75 to 89), Fair (60 to 74), Poor (40 to 59), or Critical (0 to 39)
Each of the six dimensions scored with its provenance tag: `[scan]`, `[partial-scan]`, `[heuristic]`, or `[unmeasured]`. Unmeasured dimensions show a placeholder dash, never an invented number
Issues classified as Critical, High, Medium, or Low, each with specific page URLs and the evidence behind the finding
Week-by-week fix sequence with quick wins, medium-term improvements, and strategic changes, ordered by expected impact on your score
## Targeted Dimension Audits
You don't need to run the full audit every time. Use individual dimension commands for faster, focused analysis:
| Command | What it audits |
| - | - |
| `/geo citability https://example.com` | AI citation readiness of your content |
| `/geo crawlers https://example.com` | AI crawler access via `robots.txt` |
| `/geo schema https://example.com` | Structured data detection and validation |
| `/geo technical https://example.com` | Technical SEO foundations |
| `/geo content https://example.com` | Content quality and E-E-A-T signals |
| `/geo brands https://example.com` | Brand presence on AI-cited platforms |
Each targeted command writes its own output file (e.g., `GEO-CITABILITY-SCORE.md`, `GEO-SCHEMA-REPORT.md`) so you can share individual dimension results without running the full audit.
## Provenance Tags
Every score the audit emits carries one of four provenance tags. These tell you exactly how much to trust each number:
| Tag | Meaning |
| - | - |
| `[scan]` | Scored from data actually fetched during this audit run |
| `[partial-scan]` | Some pages were sampled; others were inferred |
| `[heuristic]` | Model judgement from signals, no raw data fetched |
| `[unmeasured]` | The data source needed was unavailable; the score shows a placeholder dash |
A `[heuristic]` tag is not a failure. It means informed model judgement based on available signals. But treat it differently from `[scan]`: a heuristic score is a useful pointer, not a measured fact. When a category is `[unmeasured]`, the report emits a placeholder dash rather than a number, because a number with weak provenance still reads as hard data.
## Combining with the THL Method
The `geo-audit` skill is the first layer of the three-layer THL GEO Method:
1. **Dimensional audit** (`geo-audit`): six GEO dimensions, a composite 0 to 100 and an action plan
2. **External benchmark** (`agent-readiness-scan`): Cloudflare's independent 0 to 100, tracked as a delta across re-audits
3. **Deliverable** (`audit-report-kit`): a branded PDF and compile-checked JSON-LD from the audit data
Run the `agent-readiness-scan` skill alongside `geo-audit` so you have two independent scores: one from your own dimensional analysis and one from Cloudflare's public benchmark. On a re-audit, the delta on each is the proof of progress.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.