Draft Standard · Version 1.0 · April 2026

Artificial Intelligence
Preference Protocol

AIP-1 is a proposed open standard for declaring how your content may be used by AI systems — via ai.txt, HTML meta tags, and HTTP headers.

AIP-1 Draft Declarative AIP-1 Compliant MIT License

What is AIP-1?

As AI systems become central to the web, content creators deserve a clear, standardised way to declare their preferences. AIP-1 fills the gap that robots.txt was never designed to cover.

🤝

Neutral by Design

AIP-1 does not restrict AI — it enables transparent communication of preferences between publishers and AI developers.

🔌

Plug-and-Play Compatible

Works alongside existing standards. No infrastructure overhaul required. Add a file, a tag, or a header and you're compliant.

📐

Extensible

Designed with future AI capabilities in mind. New directives can be added without breaking existing implementations.

🌍

Ecosystem-Aligned

Built for adoption by AI labs, search engines, content platforms, and standards bodies such as the W3C.

Three-Layer Protocol

AIP-1 operates across three complementary layers to give publishers fine-grained control at every level of the web stack.

01

Root Declaration — ai.txt

A plain-text file at your domain root, analogous to robots.txt, declaring site-wide AI usage preferences.

https://example.com/ai.txt
# AIP-1 Configuration File

User-Agent: *
Allow-Training: no
Allow-FineTuning: no
Allow-Indexing: yes
Allow-Summarization: conditional

User-Agent: AcademicBot
Allow-Training: yes
Allow-CommercialUse: no
02

Document Layer — HTML Meta Tags

Per-page granular control placed in the <head> section. Overrides or supplements the root declaration.

index.html
<meta name="aip:training"   content="deny">
<meta name="aip:fine-tune"  content="deny">
<meta name="aip:summary"    content="allow">
allow deny conditional attribution-required
03

Transport Layer — HTTP Headers

Ideal for APIs and dynamically-generated content where HTML markup is unavailable.

HTTP Response Header
AIP-Policy: no-training, no-finetune, allow-index

Directive Reference

All supported directives and their accepted values.

Directive Values Description
Allow-Training yes no Use of content to build or improve AI models
Allow-FineTuning yes no Updating models using this domain-specific data
Allow-Indexing yes no Crawling and indexing content for AI search
Allow-Summarization yes no conditional Generating AI summaries or excerpts of content
Allow-CommercialUse yes no Use of content in commercial AI products

Get Compliant in 4 Steps

Implementing AIP-1 takes minutes. Follow these steps to declare your preferences.

1

Create ai.txt

Add a plain-text file to your domain root at /ai.txt. Declare your global preferences using AIP-1 directives.

2

Add HTML Meta Tags

Insert <meta> tags inside each page's <head> section for page-level overrides.

3

Configure HTTP Headers

For APIs and dynamic content, add the AIP-Policy response header to your server or CDN config.

4

Document Your Policy Publicly

Transparency is central to AIP-1. Publish a human-readable AI usage policy page for your users and partners.

Show Your Compliance

Signal to the world that your site respects AIP-1. Add the badge to your README or site footer.

AIP-1 Compliant Badge
README.md
[![AIP-1 Compliant](https://img.shields.io/badge/AIP--1-Compliant-brightgreen)](https://github.com/SquiddoTech-Systems/AIP-1/))
index.html
<img
  src="https://img.shields.io/badge/AIP--1-Compliant-brightgreen"
  alt="AIP-1 Compliant Badge"
>

Why AIP-1 Matters

Existing mechanisms like robots.txt were designed for search engines, not generative AI. They offer no way to distinguish between indexing, training, fine-tuning, or commercial use.

AIP-1 closes this gap by introducing a structured, extensible vocabulary that all stakeholders — publishers, AI developers, and standards bodies — can adopt.

View the Full Spec →
3
Protocol Layers
5
Core Directives
4
Supported Values
Use Cases