robots2.txt

robots2.txt is an AI policy file for the open web. It is meant to sit alongside an existing robots.txt file and give AI agents machine-readable instructions about what they are allowed to do with a site’s content. The page frames it as a response to a gap in robots.txt, which was built for search engine crawlers and does not address training, summarising, quoting, or building a competing product from content.

The specification defines directives for crawl, read, summarise, quote, derivative, train, store, compete, market, personalise, monetise, attribution, link-back, rate, announce, honest, content-type, editorialised, ai-assisted, primary-language, age-rating, and report-to. It also includes a chain directive for community defaults, a quick-start example, and an ask protocol for cases where a directive is set to ask. The age-rating directive supports inline path scoping, and the text says agents should check age-rating before summarising, quoting, or surfacing content to users. Agent types are also part of the model, with categories such as ai-assistant, data-harvester, ad-network, search-indexer, ai-researcher, code-assistant, and content-generator, and site owners can apply override blocks for different types.

The page is written for both site owners and AI developers. For site owners, it describes setting boundaries such as allowing summarisation while blocking training, requiring attribution, and blocking ad networks and data harvesters. For AI developers, it emphasizes one file to parse, clear directives with defined values, and no legal ambiguity. It also says the system is a shared language for the relationship between human creators and AI agents.

robots2.txt is served as a plain text file and the quick start instructs site owners to create it at the domain root. The page names version 0.2.2 and says it is open, extensible, and backwards-compatible. It does not state a pricing model or licence.