Discovery and validation · updated 6 August 2026
Machine-Readable Publishing
Machine-readable publishing makes approved public information easier to discover and interpret. It is useful infrastructure, not a guarantee of indexing, citation, ranking, or accurate AI answers.
Start with readable public pages
Publish the information you want people and software to understand on stable, crawlable HTML pages. Give each page a unique title, description, canonical URL, internal links, and relevant structured data that matches its visible content.
Validate the actual payload
A page can render normally while JSON-LD is malformed or missing from a crawler’s view. Validate the deployed structured-data payload after release, and treat validation failures as release blockers for the affected public record.
Use discovery files as supplements
Sitemaps, robots.txt, public JSON, and llms.txt make discovery intent clear. They do not grant access to protected content and should not be presented as guarantees that a search engine or AI system will crawl, index, cite, or summarize a work.
Independent corroboration remains separate
A publisher can control its own metadata and documentation. Independent reviews, citations, and mentions are separate evidence with their own editorial standards; never manufacture them or treat a self-authored record as independent confirmation.
This page develops ideas first discussed in the Agentic Publishing community. The documentation states the current public boundary and does not turn discussion or roadmap material into a product guarantee.
Related documentation
For a concrete public record, inspect the metadata-only Context Pack demo. It remains separate from protected manuscript content and private editorial systems.