Publish a Mirror

AI systems retrieve content from the web. They can only cite and verify content they can find, parse, and trust. Without a structured machine interface, your domain is invisible to AI retrieval systems — or at best scraped and guessed at.

An AIEP Mirror is the fix: a small, structured layer you publish at a known path on your domain. Five JSON files. Any web server. No backend. No registration required.

Who should publish a Mirror

Any domain that publishes authoritative content people and AI systems rely on. In practice, this means:

Sector	What you publish	What you gain
Regulators & government	Legislation, guidance, enforcement records	AI systems can retrieve structured, hash-bound versions rather than relying on unsupported scraping
Legal publishers	Case law, statutes, professional standards	AI legal tools retrieve verified versions, not outdated scrapes
Financial data providers	Rates, disclosures, regulatory filings	AI finance systems can independently verify the data they cite
Academic & research institutions	Papers, datasets, methodology documentation	Citations are hash-bound — changes after publication are detectable
Enterprise compliance teams	Policies, audit records, certifications	AI compliance tools can verify your published position without calling you
Standards bodies	Specification documents, schema definitions	Implementors get machine-readable access to normative documents
Any AI-native product	Evidence artefacts, response records, verification endpoints	Participate in federated evidence networks automatically

The minimum viable adoption — index.json and metadata.json at /.well-known/aiep/ — takes under 30 minutes. See Mirror Implementation for the full guide.

What it does for you

Without a Mirror	With a Mirror
AI retrieval systems scrape or skip your domain	AI systems discover and retrieve your artefacts directly
Your content is cited by name but not verified	Every artefact is hash-bound — changes are detectable
No standard way to assert what you publish	`metadata.json` declares your identity and capabilities in machine-readable form
Cannot participate in federated evidence networks	Other Mirrors can federate with yours — your artefacts appear in cross-organisation evidence queries

Federation: the network effect

This is the part that matters at scale. Once your Mirror is published, other Mirror operators can declare it in their federation.json manifest. An AI agent traversing their Mirror will automatically traverse yours.

What this means in practice: If a law firm, a regulator, and an industry body all publish conformant Mirrors and federate with each other, an AI agent querying any one of them retrieves evidence from all three — without any central coordination, API agreement, or data sharing contract. The path convention is the discovery mechanism.

Your artefacts appear in cross-organisation evidence queries the moment another operator federates with you. No registration. No central authority.

Mirror Federation → — how to declare federation links and traverse federated surfaces.

Publish in five steps

The minimum viable Mirror requires no backend. Add these five files to any existing web server:

/.well-known/aiep/index.json        # Discovery endpoint — lists all published artefacts
/.well-known/aiep/metadata.json     # Publisher identity and protocol version
/.well-known/aiep/version.json      # Protocol version declaration
/.well-known/aiep/canon/self-test-vectors.json  # Canon hash self-test
/robots.txt  (add line: AIEP-Mirror: allow)

A working reference implementation is live at /.well-known/aiep/ on this hub. Use it as your template.

Schema definitions and validation tooling: Downloads → aiep-well-known and aiep-hub-validator.

Validate your Mirror →

What a Mirror contains

At minimum, a conformant Mirror contains:

File	Purpose
`index.json`	Lists all published artefacts with paths, types, and hashes
`metadata.json`	Declares publisher identity and protocol version

Beyond the minimum, a Mirror can contain any AIEP artefact type: certificates, audit events, compliance records, schemas, ledger entries, and downloads. Publish as much or as little as your use case requires.

How AI systems discover your Mirror

AI retrieval systems discover your Mirror by querying GET /.well-known/aiep/index.json — the single entry point. Everything else is linked from the index.

What happens when an AI retrieval system encounters your domain:

Checks for /.well-known/aiep/index.json — if absent, no AIEP discovery
Reads metadata.json — establishes your publisher identity and schema version
Fetches artefact paths from index.json
For each artefact: validates schema conformance, checks artefact_hash
Admits artefact into retrieval results if all checks pass

No registration is required to publish a Mirror. The path convention is the discovery mechanism.

Mirror vs website

	Website	Mirror
Audience	Humans	Machines + humans
Format	HTML	JSON (schema-conformant)
Verifiable	No	Yes — hash-bound
Discoverable by AI	By crawl / guess	By convention at `/.well-known/aiep/`
Requires interpretation	Yes	Reduced — structured JSON and declared artefacts

Implementation guide

Mirror Implementation → — full technical guide: directory structure, required files, schema conformance, hash computation, and endpoint validation.

Mirror Adoption → — who should adopt a Mirror and when.

AI Retrieval · Architecture · Get Started · Validator · Downloads