LLMO
The problem
Language models and autonomous agents increasingly make consequential decisions based on claims about organizations. Who operates a domain. When a statement was published. Whether a credential is current. Which entity a communication came from. Today these claims are scattered across websites, press releases, third-party directories, and social platforms. The formats are inconsistent. The sources are not cryptographically signed. The time of assertion is usually unrecoverable.
A consumer that needs to verify an organizational claim, whether a reasoning system, a search index, or a human operator, has no canonical place to look and no canonical format to parse. The result is that organizational facts get synthesized from whatever happens to be surfaced at retrieval time, with no authoritative source to anchor them.
LLMO addresses this by defining a canonical location, a canonical format, and a verification model.
The approach
Organizations publish a signed JSON document at /.well-known/llmo.json on their primary domain. The document contains claims about the organization: identity, operators, endpoints, publication timestamps, and signatures. The schema is versioned and machine-readable. Signatures bind claims to the publisher’s control of the domain.
Consumers fetch the document, validate it against the schema, verify signatures, and use whichever claims they trust for their purpose. LLMO does not impose a trust authority. Each consumer chooses its own trust model.
The protocol is vendor-neutral, open, and intended to be widely implemented.
What exists today
Specification v0.1 is published with a JSON schema, five test vectors, and a reference validator at validate.llmo.org. A governance process is active with three accepted improvement proposals covering the process itself, core-proposal submission mechanics, and authoring conventions. OpenTimestamps anchoring is live for accepted proposals.
Where to go next
If you’re evaluating LLMO as a publisher: read the specification, validate a draft document against the reference validator, and publish at your domain’s /.well-known/llmo.json.
If you’re evaluating LLMO as a consumer: read the specification to understand the document structure, fetch documents from publishers you care about, and apply whatever trust model fits your use case.
If you’re evaluating LLMO as an implementer of tooling: the specification, JSON schema, and test vectors together define the full interoperability surface. A conformant implementation can be built from these alone.
Governance
LLMO is stewarded by Diverse.org, a California 501(c)(3) nonprofit. The specification is vendor-neutral. Governance details, the editor role, and decision-making processes are documented on the about page.