YAML
Page metadata, indentation-clean.
YAML is the lingua franca of configuration in the AI/ML stack - model cards, prompt templates, agent definitions, OpenAPI specs, GitHub Actions, and Kubernetes all use it. LLMs generate and read it fluently, so publishing config and metadata in valid YAML lets agents and AI engines manipulate your system without an HTML scraper.
Geordy generates this format for every page on your site that warrants it - auto-detected by page type, served from your subdomain, and refreshed on the cadence below.
# AI Optimization Configuration
site:
name: "Geordy AI Platform"
domain: "geordy.ai"
category: "GEO / LLMO"
ai_targeting:
primary_systems:
- ChatGPT
- Claude
- PerplexityHow to make this format work for AI.
- Always use 2-space indentation, never tabs - tabs are explicitly forbidden by the spec for indentation.
- Quote strings that look like booleans, numbers, or dates (`"yes"`, `"01:23"`, `"2026-05-12"`) to avoid implicit type coercion (the famous Norway problem: `NO` -> false).
- Use block style (`key:` newline + indent) for anything human-edited; reserve flow style (`{a: 1}`) for short inline values.
- Validate against a JSON Schema where possible (e.g. `yaml-language-server: $schema=...`) so editors and CI catch errors early.
- Trusting `yaml.load()` in PyYAML on untrusted input - it can execute arbitrary Python; always use `yaml.safe_load()`.
- Unquoted version strings like `1.10` parsed as the float `1.1`, silently breaking downstream tooling.
- Mixed indentation across a single block (3 spaces here, 4 there) - most parsers fail with cryptic errors far from the actual line.
Use cases, scored for AI visibility.
Real sites already shipping YAML.
Tools & resources for YAML.
- Human-readable syntax with minimal punctuation
- Widely supported across programming languages and platforms
- Eliminates bracket clutter through indentation-based structure
- Indentation-sensitive (error-prone)
- Slower parsing than JSON
- Security concerns with unsafe loading
Written for these readers. Which of them actually request it is measured, not assumed: see how we measure.
- YAML specification 1.2.2A human-friendly, cross-language, Unicode-based data serialization language.
Share of eligible sites where each crawler requested this file at least once, over the 90 days to 21 November 2026.
| Crawler | Evidence | Sites with a request |
|---|---|---|
| GPTBot | Emerging pattern | about 80% |
| Amazonbot | Emerging pattern | about 70% |
| ChatGPT-User | Emerging pattern | about 60% |
| OAI-SearchBot | Emerging pattern | about 50% |
| PerplexityBot | Emerging pattern | about 50% |
| Meta-ExternalAgent | Emerging pattern | about 50% |
| bingbot | Emerging pattern | about 40% |
| Applebot | Emerging pattern | about 40% |
| Claude-SearchBot | Emerging pattern | about 30% |
| Googlebot | Emerging pattern | about 30% |
| Claude-User | Emerging pattern | about 20% |
| ClaudeBot | Emerging pattern | about 20% |
| Perplexity-User | Emerging pattern | about 20% |
| Meta-ExternalFetcher | Emerging pattern | about 20% |
| CCBot | Emerging pattern | about 20% |
| Bytespider | Emerging pattern | about 20% |
| DuckAssistBot | Emerging pattern | about 20% |
| cohere-ai | Emerging pattern | about 20% |
| YouBot | Emerging pattern | about 20% |
| Claude-Web | Emerging pattern | 0% |
| anthropic-ai | Emerging pattern | 0% |
| Google-CloudVertexBot | Emerging pattern | 0% |
| FacebookBot | Emerging pattern | 0% |
| MistralAI-User | Emerging pattern | 0% |
| Diffbot | Emerging pattern | 0% |
| Timpibot | Emerging pattern | 0% |
| PetalBot | Emerging pattern | 0% |
Figures are rounded and cover only sites that served the file for the whole window and were reached by crawlers at least once in it; sites with no crawler requests at all are left out. Requests are to Geordy-served files, identified by user agent; a request is not evidence of use or citation. Pairs with too few independent sites are not shown. See how we measure.