Google search history as an interest poll

As many of you know, I follow the network neutrality discussions at an above average attention level. So it was natural for me to come across this article posted over at the Technology Liberation Front (TLF). The substance of the argument aside, what I found interesting was the use of search results as a tool to gauge public interest/sentiment. It is clearly not a random-selection type poll, but it is certainly broadly based. I wonder how well this kind of measure would compare to a more rigorous, carefully designed and implemented survey. Clearly you can't ask the same types of questions, but if search and level of interest are highly correlated, then for some kinds of questions this approach might lead to a good first approximation.