I see the Google N-gram viewer used and misused quite a lot e.g. to settle disputes with pseudo-quantitative accuracy that are actually matters of grammar, or inherently subjective and thus not "solvable". The writers (actually, the arguers on the talk page) of the N-gram article in Wikipedia needed to consult this page. Here are the underlying assumptions, in a simplified but still very useful, yet not widely distributed FAQ. Direct from the source: The developers of the N-gram project data and viewer.