IETF AI Preferences (AIPref)

In progress

Internet Engineering Task Force is working on a standardized preference signal for AI agents and crawlers ("building blocks that allow for the expression of preferences about how content is collected and processed for Artificial Intelligence (AI) model development, deployment, and use.")

Visit websiteBack to browse
Website
datatracker.ietf.org/wg/aipref/about
Latest tracked evidence
May 01, 2026Attachment draft listed as expired
Last checked
Jul 15, 2026
Profile updated
Jul 15, 2026
Primary approach
Preference signal
Practical force
Voluntary signalIt only affects actors that detect and honor the signal; it does not itself prevent reuse.
Pipeline
Collect / Train / Retrieve
Status rationale
Publicly documented, but still emerging or not fully deployed.

Enforcement

Voluntary preference signal. Communicates a requested condition or restriction to automated systems.

It only affects actors that detect and honor the signal; it does not itself prevent reuse.

Catalog status describes public availability, not legal validity, adoption, or proven effectiveness.

What it is

The AIPref effort investigates a standardized way to express preferences to AI systems, including crawlers and agents. The working group is still active, with current vocabulary work and an expired attachment draft in its public record. No production deployment signal is available yet.

IETF AI Preferences (AIPref) publishes or proposes a machine-readable preference signal for web content across the collection, training, and retrieval stages. The signal communicates requested conditions; compliance depends on discovery, interpretation, and voluntary support by downstream systems. Public materials describe an in-progress proposal or implementation; the newest dated source in this profile is “Attachment draft listed as expired” (May 1, 2026). These details describe the published mechanism and evidence, not a finding about legal validity, adoption, or effectiveness.

Limitations

The AIPref working group remains active, but the attachment draft referenced by the HTTP header and robots.txt examples is currently expired on the IETF Datatracker. Treat those snippets as draft-history examples until newer attachment work replaces them. The working-group charter explicitly excludes technical enforcement: AIPref communicates preferences and does not itself block access or compel compliance.

Evidence trail

Adoption signals

No public adoption figure found.

Examples

HTTP response header draft exampleSource

The expired attachment draft showed AI preferences delivered directly in the HTTP response.

HTTP/1.1 200 OK
Content-Type: text/plain
Content-Usage: train-ai=n
robots.txt draft exampleSource

The expired attachment draft also sketched a domain-level robots.txt expression.

User-agent: *
Allow: /
Content-Usage: train-ai=n
HTML meta-tag draft exampleSource

Earlier vocabulary work also explored per-page HTML metadata for finer-grained preferences.

<meta name="AI-PREF" content="allow_training=false; retention_period=0">