Normally clicks have a "gclid" query param. Google is introducing 2 new query params to somehow attribute clicks using modeling + machine learning (somehow).
Edit: here's a detailed description of how Google is attempting to track conversions using machine learning. I have no idea how this could possibly work without some kind of fingerprinting or user profiling or IP address. Almost feels like "modeled conversions" powered by ML is a way to do fingerprinting without explicitly having an algorithm that blatantly uses fingerprinting.
https://support.google.com/analytics/answer/10710245?sjid=85...
Edit 2: The new query params are "gbraid" and "wbraid". Googling those turns up more details.