We also have to consider the (near future?) implications of machine learning on sifting through the "noise" that is generated with tools like this. In fact, by

We also have to consider the (near future?) implications of machine learning on sifting through the "noise" that is generated with tools like this. In fact, by knowing that one is using such a tool, one could envision a scenario where this data adds more information about the user and not less - in particular about everything that one is not actually searching for or what OS the user is actually using. Consider that just clicking on a random search link with a browser automation engine probably doesn't look like a real user browsing without considerable effort on the developers part. Here we arrive at counter ML :)