Glossary Term

llm.json

A small JSON file at the root of a website that tells a language model what the site is, what it publishes, and where the rest of its machine-readable files live.

llm.json is a one-file summary of a website written for AI systems. It names the site, describes it in a sentence, lists its main machine-readable files, and points to its sibling sites. It is served at /llm.json and again at /ai/llm.json.

The Digital Karma Federation retired llm.json in v8.0 because the manifest already carried the same facts. Crawler logs disagreed. Over the 35 days ending September 22, 2026, classified AI crawlers requested llm.json 205 times across the portfolio and the manifest 115 times. v8.1 brought it back as a file that is generated every night from the manifest and never written by hand. The full story is here.

In Practice

This site publishes its own at /llm.json. It lists the site name, a one-sentence summary, the manifest and catalog addresses, and every sibling site in the constellation.

Worth Knowing

The request counts show crawlers asked for the file. They do not prove any crawler used it. The reason to publish it anyway is cost: a small generated file costs nothing, while a 404 or a stale copy tells the crawler nothing true.

Related Terms

llms.txt

A short markdown file at the root of a website that introduces the site to a language model and links to the pages worth reading first.

Digital Karma Federation

The open standard every site in this portfolio uses to publish machine-readable files about itself, so AI systems can read the site without guessing.

Structured Data

Machine-readable markup, like JSON-LD, that states what a page or entity is directly, instead of leaving a search engine or AI system to infer it from prose.

← Back to Glossary