ai.txt — agent readiness

Why an agent cares

It states how the site wants to be credited when an AI system uses its content, which no other file covers. Adoption of the lineage we validate is effectively zero, so a site publishing one has opted in deliberately and is worth getting right.

Adoption

Specification only. A published specification or draft. Deployments are limited to the authors' own demos.

Direction of travel: ► flat, as at 2026-07-28.

The Geodocs spec names no adopters. Originality.ai counted 397 ai.txt files across 3M+ tracked sites (May 2026), mostly the older Spawning opt-out variant.

What we check

Absence is never a finding. Every check below runs only once aiTxtPresent is true, so a site that has not adopted this is not assessed on it and appears in no report. 3 of these 3 can fail it. In a Lumar crawl these land on one row per site, in the crawl-level table where itemType is access.

Check	Severity	Raises	On whose authority
`aiTxtNotKeyValue`	high	usability	Geodocs profile
`aiTxtInvalidSourceUrl`	medium	usability	Geodocs profile
`aiTxtUnknownAttributionValue`	medium	usability	Geodocs profile

Examples

Both of these are fixtures the test suite runs through the real collector, not snippets written for a document: the first is a shape we score as valid, the second is one we flag.

correcttexttest/e2e/test-pages/access/correct.ai.txt

# ai.txt — Geodocs AI Agent Content Spec (https://geodocs.dev/ai-agents/content-spec)
# Correct fixture: present and well-formed → aiTxtPresent + aiTxtValid true, no defects.

# aiTxtNotKeyValue (false): every line is a "Key: value" directive.
User-agent: *
Allow: /

# aiTxtUnknownAttributionValue (false): Attribution-required is one of yes/no.
Attribution-required: yes

# aiTxtInvalidSourceUrl (false): Source-url is an absolute http(s) URL.
Source-name: Acme Corporation
Source-url: https://acme.example

Citation-format: [Title] — [Source-name] ([Source-url])

present but wrongtexttest/e2e/test-pages/access/issues.ai.txt

# ai.txt — Geodocs AI Agent Content Spec (https://geodocs.dev/ai-agents/content-spec)
# Issues fixture: present but wrong → aiTxtPresent true, aiTxtValid false.
# The file IS "Key: value" (aiTxtNotKeyValue stays false — a prose file is covered by unit tests);
# individual directive values are malformed.

User-agent: *
Allow: /

# aiTxtUnknownAttributionValue: Attribution-required must be yes/no, not "sometimes".
Attribution-required: sometimes

# aiTxtInvalidSourceUrl: Source-url must be an absolute http(s) URL, not a relative path.
Source-name: Acme Corporation
Source-url: /company/about

Citation-format: [Title] — [Source-name]

Specifications

Document	Revision	Kind
Geodocs AI Agent Content Spec	`May 2026`	specification

Last re-read against the published documents: 2026-08-12.