If you would like to handle the freshness state of the document so the latest version is the one that comes up as part of the engine itself (in AI Search), we're currently working on a feature improvement that takes care of this functionality so this gets translated to the answers returned to the LLM when performing RAG.