ai.txt — A Policy File Still Taking Shape

The name ai.txt has been used for more than one attempt to let websites express preferences about AI use. An early proposal promoted by Spawning in 2023 placed an ai.txt file at the root of a website, principally to communicate preferences concerning text-and-data mining.

A newer proposal, published in June 2026 as an independent Internet-Draft, describes /.well-known/ai.txt and a JSON equivalent, /.well-known/ai.json. It proposes signals covering activities such as training, inference, scraping, indexing and caching, together with licensing, attribution and user-agent-specific conditions.

These signals remain advisory. They express a website operator’s intentions but do not enforce them technically or create agreement about their legal effect.

The 2026 document uses the IETF Internet-Draft process, but it is not an IETF standard and has not been adopted as a working-group document. The IETF’s AI Preferences working group is pursuing a related but separate standardisation effort involving common preference vocabulary and ways to attach it to content.

Adoption of ai.txt by AI providers is presently unclear. The major provider documentation reviewed for this project describes robots.txt controls, but does not make a comparable commitment to discover or honour ai.txt.

Because its location, syntax and status are still developing, any experiment with ai.txt must state precisely which version and endpoint it publishes. A request can demonstrate discovery of that endpoint, but not acceptance or application of the policy it contains.