llm.json is a public JSON file that gives AI a compact record of what a website is, what it publishes, and which URLs deserve attention. It does not replace the pages. It gives AI a better place to begin checking them.
On this site, llm.json is generated instead of hand-written
AI to AI Websites builds /llm.json from the site's current manifest, Dataset catalog, health record, services, and page inventory. The build also publishes the JSON at /ai/llm.json and writes a plain text reading guide at /llms.txt.
That distinction matters. Editing one summary by hand feels easy until the site changes. Generating the files from the current public records makes stale names, missing URLs, and conflicting counts harder to hide.
The llm.json contract has three jobs
- Identify the website, its purpose, its publisher, and its preferred origin.
- Name the important topics, services, datasets, pages, and public endpoint files.
- Give AI stable URLs it can retrieve when it needs the full page or source record.
A valid JSON file can still fail all three jobs. The useful test is whether the names and links match the live website a person can inspect.
What to check before trusting an llm.json file
- The declared website name and origin match the canonical pages.
- Important URLs return the page or file the record describes.
- The services and datasets exist in the visible website instead of only in the summary.
- The update and health information reflects the current build.
- The JSON summary and text guide agree without pretending they are the same format.
Open the live llm.json files
- Open /llm.json Generated JSON summary at the site root.
- Open /ai/llm.json AI directory mirror of the same generated record.
- Open /llms.txt Plain text guide to the site.
- Inspect all public resources Guided links to the manifest, catalog, content records, and public indexes.
What llm.json does not prove
A public llm.json file proves that the file can be retrieved. It does not prove that an AI system used it, cited it, trained on it, or trusted every claim inside it. Those are different questions and need their own data.
The file earns its place when it stays current, points to real information, and makes the website easier to inspect. If it drifts away from the live pages, it becomes one more conflicting answer.