curious how this differentiates from just adding ai-specific directives to standard robots.txt. like do you have a use case where you need the age rating that simple disallow rules cant cover
Good question — the short answer is that robots.txt only understands Allow/Disallow. It's binary: a bot can go here, or it can't.
robots2.txt adds a permission layer that robots.txt was never designed for. For example, you might want an AI assistant to read and summarise your blog posts (so users can discover your content), but not train on them, not store them permanently, and not build a competing product from them. That's four separate permissions on the same path — robots.txt can't express that.
Other things you can't do with robots.txt: require attribution when your content is quoted, distinguish between an AI assistant and a data harvester, set a "session-only" storage policy that explicitly prohibits vector stores and RAG indices, or use the ask protocol to grant one-time permissions dynamically.
The age-rating is just one of 20+ directives. The real value is giving site owners granular control over what AI agents do with their content, not just where they can go.