What if an AI service uses cached or previously indexed information and does not fetch the site again?

That can happen. An AI platform may cite or use information without making a fresh server request at the exact moment a user asks a question. In that case the Digital Twin cannot record a request that never reached its infrastructure. This is one reason visible citations, first-party retrieval evidence and confirmed referrals are kept as separate signals instead of pretending any one system sees everything.