I think part of the interest for students isn't so much gaining info they care about as much as seeing patterns (possibly unexpected patterns) and trying to figure out a "story" behind them. The search topic could be something as mundane as a search term "pumpkin' and noticing the search term has a regular cycle of spikes (when viewing 5 year data). Upon closer inspection, it seems to be annual... around the same time each year (I wonder why)... and always followed by a smaller spike (around the end of November hmmmm)...
Sometimes searching for items in pop culture (say "iphone 7") may give predictable results (spike around the release date) that student may still find interesting that actual real-life data from around the world corresponds to what they might predict.
Also, if you wanted to take a detour... "Why doesn't China search for anything?" :oP
I feel a crucial part of this lesson is distinguishing between observations and inferences. I did this lesson last year several days before the election and the front page of Google Trends showed that across the country the search term "trump" dwarfed the search term "clinton". Many of my students reacted "omg Trump is beating Clinton" and after hearing groups of students comment on this, I paused everyone to ask them what the data actually represents. "Is it saying that all these people are voting for Trump?" It took them a few seconds to think about it realize what the data actually represents is how many searches are "trump", and anything outside of that is an inference - maybe there's more searches because he's more controversial or interesting, etc. (Although it might not be a great example to bring up today, given the actual outcome may just feed students' preconceived notions...)