Open methodology
Agent-Readiness Methodology
The deterministic checks Thally uses to evaluate whether documentation can be discovered, parsed, understood, and cited by AI systems.
Methodology version 1.0. Last reviewed July 19, 2026.
What the score measures
The score is a conformance result, not a prediction that an answer engine will cite a page. It evaluates the technical conditions that let a machine find the canonical source, retrieve useful content, understand its structure, and verify its freshness.
Evaluation categories
- Discovery: sitemap, robots, llms.txt, feeds, and explicit alternate formats.
- Identity: canonical URLs, language declarations, titles, descriptions, authors, and dates.
- Structure: one descriptive H1, logical headings, semantic HTML, JSON-LD, and API schemas.
- Retrieval: useful Markdown or JSON, stable URLs, server-rendered content, and successful responses.
- Trust: cited primary sources, editorial ownership, correction paths, and claim freshness.
- Parity: consistent claims across HTML, metadata, schema, feeds, and machine-readable files.
Release checks
The website release audit fails for unsafe production origins, malformed discovery files, missing critical metadata, invalid JSON-LD, duplicate canonical URLs, or prohibited copy characters. Warnings identify title or description lengths and optional trust signals that need editorial judgment.
Limitations
No technical score guarantees ranking, inclusion, or citation. Answer engines use their own indexes and policies. Real visibility must be measured separately with stable prompts, cited-source capture, analytics, and search-platform data over time.