Discoverability — Declaring Your Site to Agents

The agent-readiness dimension that declares your site to AI agents with llms.txt, sitemaps, robots and agents.json so they can find you before reading you.

What the discoverability dimension means

Discoverability is the first agent-readiness dimension: it makes an AI agent able to find your site and its most important pages before it does anything else. It is the Layer-1 declaration work — the machine-readable files an agent fetches at a known path to learn what you publish, where it lives, and how it is organized. You cannot expose tools, serve content or take payment from an agent that cannot discover you, which is why this dimension ships first in the implementation order.

Signals and standards it covers

The discoverability dimension groups the standards that announce your site to agents at well-known locations:

llms.txt — a file at your site root that lists your most important pages as a clean, linkable index for agents (with llms-full.txt as the expanded variant).
AGENTS.md / agents.txt — agent-facing guidance and declarations, increasingly governed under shared community specs.
agents.json — a manifest that declares the agent-facing actions and capabilities your site exposes (its discovery facet; the action surface bridges to the Capabilities dimension).
XML sitemap and robots.txt — the classic crawl-control pair that still tells agents which URLs exist and which are in or out of bounds.
.well-known discovery and NLWeb — standardized discovery endpoints an agent can probe without guessing.

Each standard names a primary spec you should verify against at build — for example, the llms.txt format is defined at llmstxt.org. Treat the spec, not an internal note, as the citable source.

How the Agent-Readiness Audit scores it

The Audit runs one check per standard in this dimension and scores discoverability on whether each declaration resolves correctly at its expected path. The anchor check is discoverability.llms_txt: it passes when /llms.txt exists, returns the right content type, and links real pages. Companion checks confirm a reachable sitemap, a robots policy that does not accidentally block agents you want, and a parseable agents.json. Every check is pass/fail against a machine-verifiable criterion — the file is there and valid, or it is not — so the score is reproducible. This site demonstrates the dimension by serving its own /llms.txt and agents.json live.

← Agent-Readiness · .md