AI Preferences (AIPREF)
AIPREF declares AI access preferences. Originary makes policy outcomes provable.
Overview
Publishers declare what agents may do. Originary issues a signed record when agents comply, turning advisory preferences into enforceable, verifiable evidence.
AIPREF is an IETF Internet-Draft for expressing AI access preferences in machine-readable format. Similar to robots.txt for web crawlers, AIPREF allows websites to declare policies for AI agent access, scraping permissions, and usage terms.
Originary extends AIPREF with PEAC-Receipt verification, enabling sites to enforce policies cryptographically. Agents discover policies via /.well-known/aipref.json, and Originary issues receipts proving policy compliance.
Getting Started with AIPREF
Implementing AIPREF on your website starts with creating a machine-readable policy file at /.well-known/aipref.json. This file declares your AI access policies in standardized JSON format, allowing AI agents and crawlers to discover and respect your preferences automatically.
The AIPREF standard supports multiple policy types: scraping permissions, training data licensing, commercial use restrictions, rate limits, payment requirements, and attribution formats. Agents that support AIPREF will fetch your policy file before accessing content and adjust their behavior accordingly.
Originary reads your aipref.json configuration and maps its fields to PEAC policy checks. When an agent complies with the stated requirements (payment, attribution, usage limits), a signed receipt records the compliance. The receipt can be verified offline by anyone with the issuer's public key.
AIPREF File Format
Example /.well-known/aipref.json
{
"version": "1.0",
"ai_access": {
"scraping": "allowed_with_attribution",
"training": "prohibited",
"commercial_use": "requires_license"
},
"rate_limits": {
"requests_per_hour": 100
},
"payment": {
"required": true,
"endpoint": "https://api.example.com/payment"
},
"attribution": {
"required": true,
"format": "Source: example.com"
}
}Use Cases
Content Licensing
Publishers declare licensing terms for AI training data. Agents discover terms and obtain licenses before scraping.
Rate Limiting
Sites declare acceptable request rates. Agents respect limits to avoid being blocked or incurring overage fees.
Attribution Requirements
Content creators specify attribution formats. Agents include proper citations when using content.
Implementation with Originary
Originary reads your aipref.json file and enforces policies at the edge. When agents comply with policies (e.g., by paying required fees), Originary issues PEAC-Receipts that prove compliance.
Agent Request with Policy Compliance
GET /content HTTP/1.1 Host: example.com PEAC-Receipt: eyJhbGc...policy-compliance-proof User-Agent: MyAgent/1.0
Benefits for Publishers
- Machine-readable AI access policies
- Verifiable enforcement of licensing terms
- Automated compliance verification
- Monetization of AI training data
- Standards-based approach compatible with IETF specifications