AI Domain Data Standard – Self-Hosted Domain Identity for AI Agents

TL;DR

TL;DR: The AI Domain Data Standard lets any domain publish a self-hosted JSON profile at /.well-known/domain-profile.json that AI systems, search engines, and automated agents can consume — eliminating hallucinated domain information without a third-party dependency.

Source and Accuracy Notes

• Project page: ai-domain-data.org
• Source repository: github.com/ai-domain-data/spec
• License: MIT (verified via repository LICENSE file)
• Specification: ai-domain-data.org/spec — current version v0.1.1
• HN launch thread: Not confirmed via YC batch — this is an open-source community standard, not a YC-funded company

What Is the AI Domain Data Standard?

When an AI system crawls your site, it infers who you are from your content, meta tags, and third-party databases — a process prone to outdated information and outright hallucination. The AI Domain Data Standard (AIDD) solves this by letting any domain publish a canonical, self-hosted JSON record that machines can read directly.

The record lives at https://yourdomain.com/.well-known/domain-profile.json and optionally mirrors via a DNS TXT record at _ai.yourdomain.com. Both are read in preference order, giving resilient fallback coverage.

The format is deliberately simple — a minimal valid record requires just four fields:

{
  "spec": "https://ai-domain-data.org/spec/v0.1",
  "name": "Example Site",
  "description": "A minimal valid AI Domain Data record.",
  "website": "https://example.com",
  "contact": "[email protected]"
}

Optional fields include logo, entity_type (using schema.org @type values like Organization, Person, Blog), and a raw jsonld block for embedding complete schema.org JSON-LD.

The standard is built on schema.org vocabulary, so it slots directly into existing structured-data pipelines. The specification is vendor-neutral, MIT-licensed, and has no fees or lock-in.

Setup Workflow

Step 1: Generate a domain profile

Use the online generator at ai-domain-data.org/generator — no account required. The tool validates all required fields, lets you preview the JSON output, and provides ready-to-use DNS TXT record values.

Alternatively, install the CLI:

npm install -g @ai-domain-data/cli
aidd init --domain example.com

This creates a domain-profile.json file in the current directory, pre-filled with your domain’s information.

Step 2: Publish the record

HTTPS method (recommended, authoritative):

Upload domain-profile.json to your web server so it is accessible at https://yourdomain.com/.well-known/domain-profile.json.

DNS TXT fallback method:

Publish a TXT record at _ai.yourdomain.com with the value:

ai-json=<base64-encoded JSON>

For example, using Cloudflare:

# Encode the JSON
DOMAIN_JSON=$(cat domain-profile.json | base64)

# Add TXT record via Cloudflare API
curl -X POST "https://api.cloudflare.com/client/v4/zones/${ZONE_ID}/dns_records" \
  -H "Authorization: Bearer ${CLOUDFLARE_TOKEN}" \
  -H "Content-Type: application/json" \
  -d "{\"type\":\"TXT\",\"name\":\"_ai.yourdomain.com\",\"content\":\"ai-json=${DOMAIN_JSON}\"}"

The spec uses HTTPS as the primary source and DNS as a fallback for systems that cannot access HTTPS endpoints.

Step 3: Verify visibility

Use the visibility checker at ai-domain-data.org/checker — enter your domain and the tool will fetch the record via both HTTPS and DNS, validate it against the JSON schema, and report any errors.

From the CLI:

aidd validate --domain example.com

Platform Integrations

Official plugins handle generation and publishing automatically for common platforms:

| Platform | Package | Link | |---|---|---| | Next.js | @ai-domain-data/nextjs | npm | | Jekyll | jekyll-ai-domain-data (Ruby gem) | RubyGems | | WordPress | AI Domain Data plugin | WordPress.org |

All integrations generate the domain-profile.json during the site build or save process, validate against the schema automatically, and handle deployment.

Why It Matters

AI hallucination about domains is a real, recurring problem. When Perplexity or Copilot cites a company, it often pulls from outdated Crunchbase data, misremembered press releases, or inferred context. AIDD gives domain owners a direct, machine-readable channel to publish ground truth — name, description, contact, entity type — that persists independently of third-party databases.

The standard is complementary to robots.txt (which tells crawlers what to avoid) and sitemap.xml (which helps discovery). AIDD answers the identity question: who is this domain?

FAQ

Q: Do I need both HTTPS and DNS methods? A: No — HTTPS is authoritative and sufficient on its own. DNS serves as a fallback for systems that cannot make HTTPS requests.

Q: Is this compatible with schema.org structured data? A: Yes. The optional entity_type field uses schema.org @type values directly, and the optional jsonld field accepts full schema.org JSON-LD blocks. The spec is built on schema.org vocabulary.

Q: Does this replace ai.txt? A: No. ai.txt tells AI systems what content they may or may not use. AIDD tells them who the domain is. They address different problems and can be used together.

Q: What happens if I stop publishing the record? A: Nothing negative — it is completely optional. AI systems and agents will fall back to inferring your domain’s identity from other sources.

Q: Is the tooling free? A: Yes. The specification is MIT-licensed, all tooling is open source, and there are no third-party fees or SaaS dependencies.

Conclusion

The AI Domain Data Standard is a lightweight, well-designed solution to a specific problem: AI hallucination about who a domain belongs to. With four required fields, a self-hosted JSON file, and official integrations for Next.js, Jekyll, and WordPress, it takes minutes to publish and requires no ongoing maintenance.

If you run a domain that AI systems interact with — an open-source project, a SaaS product, a developer tool — publishing a domain profile is one of the highest-signal, lowest-effort things you can do for data accuracy in the AI era.