The taxonomy distinguishes real-time retrieval citations from memorized parametric answers, with a requirement to surface freshness timestamps when both sources appear in one response.
For publishers, the near-term impact is cleaner bot user-agents in raw logs: fewer opaque “assistant” strings and more explicit retrieval roles you can map to rate limits.
Legal teams should pair the taxonomy updates with robots and terms review rather than blocking wholesale—selective disallow often increases low-quality summarization from unofficial mirrors.