There appears to be some lexical pollution in the identification of the searches mainly coming from the fact that “si” means both “yes” and “oneself”. This is the reason I looked at perche votare no vs. perche votare si searches which show a slight lead for NO but nonetheless a loss of traction.
I tried cleaning these up as much as I could: votare no -si -come -sì -dove -puo vs. votare si -no -vales -come -può -patente -dove -puo. While the level changes and favors the NO vote when I do so the one thing that remains constant is the loss of traction for the NO searches down the stretch. Nonetheless the levels change significantly so as to lean towards inconclusive.
I might need to look at the linguistics of this more carefully at a later point. The best identified searches for now seem to be “perche votare no” and “perche votare si”. They record a loss of traction for the NO vote and show just a slight lead which however is too close to call.
In fact if I quote these searches “perche votare no” and “perche votare si” so as to look at the exact expression I get a deadlock!