ai.txt is a robots.txt-style file in which a site declares whether its content may be used to train AI models. It was introduced by Spawning in 2023, whose API passes the permissions on to partners. A separate individual IETF Internet-Draft (draft-car-ai-txt-wellknown) proposes a /.well-known/ai.txt policy file and acknowledges Spawning’s prior use of the name.
ai.txt
ai.txt borrows robots.txt syntax and applies it to file types, so a site can say, for example, do not train on its images. Spawning describes it as a simple way to declare text-and-data-mining permissions rather than a formal standard, and compliance is voluntary. Two things now share the name: Spawning’s 2023 file, and an individual Internet-Draft, draft-car-ai-txt-wellknown, that defines a /.well-known/ai.txt with fields such as Training, Scraping and an SPDX-based Training-License. The draft has not been adopted by a working group; the IETF AI Preferences group carries preferences in robots.txt instead.
Business layer
The API Commons ai-preferences schema records an ai.txt as one source of a provider’s AI usage preferences (pointer type AITxt). See The Technical Side Translated. The Business Side Did Not.
Industry: cross-domain
