Standard · Discover
IETF AI preferences (aipref)
A shared vocabulary for how content may be used by AI, in a header or robots.txt.
What it is
The working group is defining common terms such as train-ai and search, and two ways to attach them: a Content-Usage line in robots.txt and an HTTP header. The goal is one vocabulary that every AI provider reads.
An analogy
At a glance
- Owner
- IETF aipref working group
- Maturity
- Watching The vocabulary is at draft -07 (August 2026). Its first last call did not reach consensus, and the draft that attaches it to robots.txt or HTTP headers is behind. Not final.
- Where it lives
- IETF aipref working group
- Framework domain
- Discover: Be found
- What it serves
- Say how your content may be used in a form every provider could read
Who should care
The roles this standard touches most. See all the roles on the personas page.
- Legal, risk and security. Set the policies: which agents may do what, identity, data, disclosure and liability.
- Engineering and product. Build the access, data, endpoints and actions that agents use, on open standards.
How to adopt it
- 1
Follow the working group and read each new draft.
- 2
Keep your content-use policy written down in plain words now.
- 3
Plan to map that policy to the final vocabulary when it is published.
The Institute’s view
This is the likeliest long-term home for usage preferences. It is not final and it is not yet read by anyone we can confirm. Do not build on a draft.
Standards change fast. Last reviewed November 2026.