---
language: en
title: "AI Usage Preferences"
description: "AI usage preferences express how content may be processed. Learn about the IETF’s work, its limits, and how it differs from blocking access."
canonical: "https://tropk.ai/en/resources/glossary/ai-usage-preferences"
last_modified: "2026-10-12T16:10:58.826Z"
---

# AI Usage Preferences

Reviewed on 2026-10-07.

## What it means

AI usage preferences are statements about the desired or refused uses of content by automated systems. The IETF’s AIPREF group is working on a vocabulary and ways to associate these statements with files. As of the October 2026 consultation, the vocabulary was in Internet-Draft 08, still a work in progress and subject to revision.

## Why it matters for marketing

An editorial team may want to make its articles easier to discover without authorizing every use by generative systems. Expressing this intent helps communicate the policy to operators. For marketing, the decision involves identifying desired uses and requesting a compatible implementation, while avoiding presenting a published preference as proven technical protection against collection or reuse.

## Example in practice

Imagine an agronomy magazine that publishes preferences to allow search with links to its articles and refuse generative training beyond that purpose. In a hypothetical test, the automated client recognizes the search declaration but does not interpret the training one. The magazine decides to correct compatibility for that declaration before announcing that its policy is being understood by recipients.

## How to apply or verify

The technical team should check the vocabulary version, the mechanism that associates the preference with the article, and the recipient’s declared support. An interpretation test should record the file served and the values recognized. Then, verify the service’s behavior separately; recognizing the intent does not show that it was followed when processing the content.

## Important distinctions

AI preferences, in this context, are not personal preferences in a [prompt](/en/resources/glossary/prompt). They are also not equivalent to authentication, authorization, or billing. The AIPREF scope itself distinguishes expressing a preference from enforcement mechanisms. A firewall rule can block a request; a declaration communicates intended uses. Neither, on its own, proves the fate of copies previously collected.

## Sources and references

- [AI Preferences (aipref) — Working Group Charter](<https://datatracker.ietf.org/wg/aipref/about/>) — IETF
- [A Vocabulary For Expressing AI Usage Preferences — draft-ietf-aipref-vocab-08](<https://datatracker.ietf.org/doc/draft-ietf-aipref-vocab/>) — IETF

## Continue a leitura

- [Content Signals](</en/resources/glossary/content-signals>)
- [Really Simple Licensing](</en/resources/glossary/really-simple-licensing>)
- [robots.txt](</en/resources/glossary/robots-txt>)

## About the author

[Vinicius Pacheco](</en/team/vinicius-pacheco>)

Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)