because the data is useless if not aggregated. Does the government really want to sift through billions of individual search queries with tons of metadata attached to each? No, they want to see trends from which conclusions can be drawn.
Anonymization is partly a result of aggregation, but also I can't see how and PII is relevant to the case here.
However, the government may ask for all the unaggregated search data so they can perform statistical analysis on it themselves... and "coincidently" gain access to a huge trove of personally identifying data. This is the US government we're talking about here - they do have a history of overreach in this area.
I can't imagine proof (or the worth of proof) for something imagined.