Forgive my language but I'd expect people here to understand that's horeshit, they absolutely have enough data and patterning to de-anonymize the data. They spent time making it look anonymous.
De-anonymization would require linking that click stream with something that identifies the user, rather than the click stream. Perhaps that exists, perhaps not. GP didn't provide enough material to go that far.
Exactly because Google has these powers is why they have internal processes to avoid it. Of course there are products that use non-anonymous data, but this idea that everything at Google flows around with user-IDs for everyone to use and abuse is a weird stance. Google has a lot of internal auditing and validation systems when e.g. reading logs and doing feature extraction.
But I also got way more respect for Google's internal systems after I worked there than before, so I understand your scepticism.