agenticweb.wiki

AIPREF Vocabulary AIPREF

Permit The AIPREF Vocabulary draft is the second half of the IETF AI Preferences Working Group's pair of documents. Where the Content-Usage Attachment draft defines where preference values live (robots.txt directive or HTTP header), this draft defines what those values mean: a controlled vocabulary for declaring restrictions or permissions on how digital assets may be used by automated processing systems [1]. The two are meant to be read together and are cross-referenced explicitly in the attach draft [2].

Overview

It does not define an attachment or discovery mechanism of its own — the vocabulary is inert without the Content-Usage attach draft to carry it — and it does not cover licensing, payment, or compensation terms, which are out of scope for the working group's charter as currently drafted.

As of this verification (2026-09-06), the current version is draft-ietf-aipref-vocab-07, dated 2026-08-19, matching the same publication date as the -05 attach draft, suggesting a coordinated release [1]. The document is unusually explicit about its immaturity: its own 'Note to Readers' states that presence or absence of any particular text does not indicate working-group consensus, and it is published solely as a basis for further discussion [1]. This is a stronger caveat than the boilerplate carried by most IETF drafts.

Governance is the IETF AI Preferences Working Group, with development on GitHub (ietf-wg-aipref/drafts) and discussion on the ai-control@ietf.org list. No implementer of the AIPREF vocabulary specifically (as opposed to the broader Content-Usage attachment surface) was identified in this verification pass; adoption is unmeasured. What is unresolved: the exact set of controlled vocabulary terms was not enumerated in this pass and should be checked against the rendered draft text at the GitHub Pages mirror (ietf-wg-aipref.github.io/drafts/draft-ietf-aipref-vocab.html) in a future review; whether the vocabulary will converge with, or remain distinct from, the pre-existing Cloudflare Content Signals Policy vocabulary (search/ai-input/ai-train) is also unresolved, since the two were developed independently and target overlapping use cases.

Discovery and probe

Not independently discoverable; values are carried by the Content-Usage attachment mechanism (robots.txt directive or HTTP header)

GET {origin}/robots.txt
parse:     Parse any Content-Usage directive value against the AIPREF controlled vocabulary tokens defined in this draft
on absent: No AIPREF-vocabulary token found in the attachment surface; cannot distinguish 'no preference stated' from 'publisher uses a different vocabulary (e.g. Cloudflare Content Signals)'

Live check run by the MCP probe: GET /robots.txt. Vocabulary tokens are not validated; a non-empty Content-Usage value is the signal.

{
 "path": "/robots.txt",
 "expect": {
  "text_match": "(?im)^content-usage:\\s*\\S"
 },
 "note": "Vocabulary tokens are not validated; a non-empty Content-Usage value is the signal."
}

History

Instances

No instances recorded yet.

See also

References

  1. draft-ietf-aipref-vocab-07, dated 2026-08-19, defines a vocabulary for expressing how automated processing systems may use digital assets, and states the document does not yet reflect working-group consensus. — https://datatracker.ietf.org/doc/html/draft-ietf-aipref-vocab-07 (2026-08-19) VERIFIED
  2. The vocabulary draft is developed in the same ietf-wg-aipref/drafts GitHub repository and mailing list as the Content-Usage attachment draft, confirming the two are companion documents. — https://www.ietf.org/archive/id/draft-ietf-aipref-attach-05.html (2026-08-19) REPORTED

JSON · Markdown