DNS-based AI discovery: an experimental pattern

DNS-AID is an informal label, not an adopted standard. Evaluate DNS-based agent discovery, its constraints, and safer established alternatives.

Published 2026-07-04
·
Updated 2026-07-22
·
1 min read

DNS-AID

DNS-AID is an informal label used for the idea of DNS-based AI discovery. As of this review, it is not an adopted IETF standard, a registered DNS record type, or a discovery mechanism that major AI providers publicly require.

This is not yet a widely adopted web standard. Treat it as an agent-readiness pattern, not as a guaranteed ranking factor.

How It Could Work

A private experiment might publish a DNS TXT record that advertises a discovery endpoint:

_ai.example.com TXT "aid=https://example.com/.well-known/ai-discovery.json"

This example is illustrative; the _ai label and aid= key are not registered conventions. A client would have to implement the same private agreement before the record had any effect.

Design constraints

  • TXT records have practical size, quoting, and fragmentation constraints.
  • DNS responses are cached according to TTL, so policy changes are not immediate.
  • DNSSEC can authenticate DNS data, but it does not make the linked HTTPS document trustworthy by itself.
  • A record exposes information publicly and may be collected by passive DNS systems.
  • Multiple discovery documents can drift and create contradictory instructions.

SEO Guidance

Use DNS-AID only as a supporting signal. If you adopt it, keep the linked discovery document consistent with:

Avoid putting sensitive URLs or secrets in DNS. DNS is public infrastructure.

Prefer established discovery first

Before adding an experimental DNS signal, make the canonical website and API surfaces work:

  1. Use crawlable HTML, robots.txt, sitemaps, canonical URLs, and clear navigation for public content.
  2. Use standardized well-known resources where a protocol defines them.
  3. Advertise typed resources with HTTP Link headers only when a consumer is known to understand the relation.
  4. Publish an OpenAPI document or official protocol metadata for callable APIs.
  5. Document ownership, version, TTL, rollback, and a consumer test for any DNS experiment.

Acceptance test

Do not ship a DNS-AID record merely because it resolves. Name the consuming client, capture a successful discovery trace, verify DNSSEC and HTTPS behavior where required, and test stale-cache and endpoint-failure cases. If no real client consumes it, the record is documentation overhead rather than an agent-readiness feature.

Official Resources

There is no broadly accepted official DNS-AID specification yet. The stable foundations are:

Privacy & Cookies

We use cookies to enhance your experience. By continuing to visit this site you agree to our use of cookies.