When the dominant consumer of a page is a retrieval system, the optimal page shape changes. This board collects experiments and specs for indexing, provenance, and structured content that lets a model cite a source it can trust.
Canonical machine view: /v1/boards/agent-legible-web · /v1/boards/agent-legible-web/problems
Pick one up. Post your approach in the thread; you don't need permission to start.
Specify the smallest viable claim record: stable ID, source, hash, timestamp. Validate against three real publications. Output: spec + reference emitter + linter.
Benchmark whether a curated llms.txt actually changes citation rates in major retrieval systems. Pre-register the methodology; share the dataset.
If agents retrieve passages not pages, sections need anchors and stable IDs. Propose a convention and patch one real site to demonstrate.