Improve Website AEO/GEO
The web is no longer read only by humans. AI agents — ChatGPT, Claude, Perplexity, Google AI Overviews — are now how millions of people discover information. These agents don't browse. They extract, parse, and decide in milliseconds whether your content is worth citing.
If your site isn't optimized for AI, you're invisible to a growing share of your audience.
This is an agent skill that audits your website's codebase and makes concrete improvements to boost your AEO (AI Engine Optimization) and GEO (Generative Engine Optimization) scores.
Check your score at aeo-audit.sh — then use this skill to improve it.
Install
Clone the repo, then symlink or copy improve-aeo-geo/ into either ~/.codex/skills/ or ~/.claude/skills/. See the root README for installation examples.
Usage
Once installed, invoke the skill by name:
/improve-aeo-geo
Or reference it in conversation:
"Use the AEO skill to optimize this site" "Run the AEO/GEO audit on my codebase"
The skill will:
- Detect your framework (Next.js, Nuxt, Astro, SvelteKit, WordPress, Hugo, Jekyll, and more)
- Audit your code against 16 foundational checks and 6 content quality dimensions
- Make targeted code changes in priority order
- Recommend verifying your score at aeo-audit.sh
What It Checks
Foundational (16 checks, 134 points)
| Check | Points | Pass Criteria |
|---|---|---|
| AI bot access (robots.txt) | 12 | 9 AI crawlers not blocked |
| Text depth | 12 | 250+ words per page |
| Page title | 10 | 10+ characters |
| Meta description | 10 | 50+ characters |
| Internal linking | 10 | 5+ internal links per page |
| Indexability | 10 | No noindex on public pages |
| llms.txt | 10 | Valid file at /.well-known/llms.txt |
| JSON-LD structured data | 8 | At least 1 JSON-LD block per page |
| Schema types | 8 | Recognized @type values |
| Canonical URL | 8 | rel="canonical" present |
| Single H1 | 8 | Exactly 1 <h1> per page |
| Open Graph | 8 | og:title + og:description |
| Image alt text | 8 | 80%+ images have alt |
| RSS feed | 8 | Discoverable RSS/Atom feed |
| Heading hierarchy | 6 | H1 → H2 → H3, no skips |
| AI meta tags | 6 | No nosnippet/noai/noimageai |
GEO Content Quality (6 dimensions)
| Dimension | What It Measures |
|---|---|
| Answer Readiness | Direct answers in opening paragraphs |
| Quotability | Self-contained, extractable passages |
| Evidence Density | Statistics, sources, citations inline |
| Content Depth | Substantive, multi-faceted coverage |
| Freshness | Publication dates, update signals |
| Structural Clarity | Semantic HTML, clean heading outline |
Why This Matters — The Research
All claims are from peer-reviewed academic research or large-scale industry studies:
| Finding | Source |
|---|---|
| Adding quotations = +41% AI visibility | GEO Paper, KDD 2024 — Princeton, Georgia Tech, IIT Delhi, Allen AI |
| Adding statistics = +33% AI visibility | Same peer-reviewed paper |
| Citing sources = +28% AI visibility | Same paper; up to +115% for smaller sites |
| 44.2% of ChatGPT citations come from first 30% of content | Growth Memo, 2026 — 1.2M AI answers analyzed |
| Articles 2,900+ words = 5.1 citations vs. 3.2 for <800 words | SE Ranking, 2025 — 2.3M pages studied |
| Pages with FAQ sections = 4.9 citations vs. 4.4 without | Same SE Ranking study |
| Content updated within 3 months = 6 citations vs. 3.9 for 2+ years | Same SE Ranking study |
| AI cites content 25.7% fresher than traditional search | Ahrefs, 2025 — 17M citations |
| ChatGPT drives 87.4% of AI referral traffic | Conductor, 2025 — 13.7K domains |
Supported Frameworks
The skill provides framework-specific code patterns for:
- Next.js (App Router) — metadata exports, JSON-LD, robots.ts, sitemap.ts
- Nuxt 3 — useSeoMeta, nuxt-schema-org
- SvelteKit — svelte:head, JSON-LD injection
- Astro — layout props, @astrojs/sitemap
- Remix — meta function, link tags
- WordPress — functions.php, Yoast/Rank Math integration
- Hugo — partials, built-in RSS, template variables
- Jekyll — jekyll-seo-tag, jekyll-feed gems
- 11ty — Nunjucks templates, eleventy plugins
- Plain HTML — direct tag implementation
Examples
- Next.js Blog: 42 → 87/100
- WordPress Business Site: 31 → 82/100
Contributing
See CONTRIBUTING.md for guidelines on adding new checks, frameworks, or research.
License
MIT — see LICENSE.