I think you're right. Google's page on how Trends works is pretty vague: http://www.google.com/intl/en/trends/about.html#10 http://www.google.com/intl/en/trend

I think you're right. Google's page on how Trends works is pretty vague: http://www.google.com/intl/en/trends/about.html#10 http://www.google.com/intl/en/trends/about.html#10 ... but it does include this bit: 9. Is the data normalized? All results from Google Trends are normalized... Remember, Google Trends shows users' propensity to search for a certain topic on Google on a relative basis. For example, just because a particular region isn't on the Top Regions list for the term ‘haircut’ doesn't necessarily mean that people there have decided to stage a mass rebellion against society's conventions. It could be that people in that region might not use Google to find a barber, use a different term when doing their searches, or simply search for so many other topics unrelated to haircuts, that searches for ‘haircut’ comprise a small portion of the search volume from that region as compared to other regions. The last phrase of the last sentence is the key: It suggests that Google's "traffic" number is a percentage of the total number of searches. (And then they display that number on a relative scale by dividing it by its own average over the time period, removing it even further from the realm of the concrete.) Ironically, right there on the Google Trends page for "web hosting" is a chart of the absolute rate with which the term has appeared in Google News stories over time. And that rate is increasing. I'm not sure who deserves the most blame here: Google for publishing graphs that are practically useless, or the original poster for failing to read the fine print. I'd tend to blame Google, for making poorly labeled graphs with a limited choice of normalizing methods (i.e. "no choice at all").