Something is sound if, and only if, it the premises are true and the conclusion necessarily follows from the premises.
If you're using the number of results returned to determine proper usage, there might be a high correlation between number of results and usage, but it's not a sound method. For a counterexample, consider the British and American spellings of words: if color had more results than colour, inferring that color is the proper word and colour is wrong would be incorrect.
If you're using the number of results to determine popularity, it's not a sound method, either. Firstly, the number is without context: to provide meaning, you'd need something else to compare it to. When you start doing that, you run into the problem mentioned above. If you're inferring that because it has over n results, it must be popular, realize that Google only provides a basic estimate that can change due to a number of different factors, including doing something as simple as browsing page 2 of the results.
Additionally, certain keywords are given undue weight due to their SEO value; a historical example is the keyword "mesolthelioma": a relatively obscure term for malignant tumor that caused by exposure to asbestos. It's particularly popular in search results because of the high settlement rate for asbestos exposure compensation cases.
Keywords like that highlight the reality of Google results: they're not the sum total of human knowledge and are subject to gaming. The only sound conclusion one can reach from the number of Google results is that Google has indexed a lot of pages that contain or relate to that keyword, not that the term is particularly popular or that it's correct.