You probably know about robots.txt. You might know about llms.txt. But there's a third file your site is missing — and it's the most important one for how AI agents understand who you are.
The problem
AI agents are reading the web constantly. When one visits your site, it has no authoritative source for:
- Who you actually are
- What your organisation stands for
- What you will never do
- What you're not (to prevent misclassification)
It infers all of this from your homepage content. Which means it's always guessing.
The three files
| File | Answers |
|---|---|
| robots.txt | What can you touch? |
| llms.txt | What should you read? |
| intro.txt | Who am I? |
They're complementary. A site should eventually have all three. intro.txt is the one that's been missing.
What intro.txt looks like
# intro.txt — Introduction Layer declaration
# Spec: https://ailattice.io/paper
name: Acme Corp
canonical_url: https://acme.com
category: SaaS / Project Management
mission: Help teams ship software without the chaos.
core_belief: Good tooling gets out of your way.
principles:
- Clarity over completeness
- Speed is a feature
- The team is the product
constraints:
- Will never sell user data to third parties
- Will never lock exported data in proprietary formats
- Will never charge for data export
what_this_is_not:
- Not a replacement for version control
- Not a reporting tool for executives
Drop it at yourdomain.com/intro.txt. That's it.
The key design decision: constraints are prohibitions
The spec is strict about one thing: constraints must be explicit prohibitions, not aspirational language.
✅ "Will never sell user data to third parties"
❌ "We prioritise user privacy"
The first is actionable. An AI agent can check whether a proposed action violates it. The second is vague enough to mean anything.
Two paths to conformance
Easy path (2 minutes):
Drop /intro.txt at your site root. Plain text. Works on any host, any stack, no code required.
Correct path (full implementation):
Compile a constitution and serve it at GET /constitution with Content-Type: application/constitution+json. This is the machine-native cognitive entrypoint — a conforming AI agent retrieves it before any entity-specific reasoning begins.
Both paths are in the spec. You can start with intro.txt and upgrade later.
It's a W3C standard
The Introduction Layer is governed by the W3C Introduction Layer Community Group — not by us. The spec is CC BY 4.0. The reference implementation (AILattice) is Apache 2.0 with full patent grant.
No single entity can lock this down. That's by design.
What's available
- Full specification (15 sections, RFC-style): ailattice.io/paper
- Developer hub with quickstart, API docs, conformance tests: ailattice.io/developers
- Public registry of certified sites: ailattice.io/registry
- GitHub (Apache 2.0): github.com/madazz01/AILattice
- Conformance test suite (16 tests, open): /conformance
Add your site
The registry listing is free. The standard is free. The conformance tests are open.
Add intro.txt to your site and submit it to the registry — takes under 3 minutes.
Top comments (0)