There are two problems with using Google to check your English:
- The Web is full of bad English written by non-native speakers and verbally incompetent native speakers. This is a serious problem when you’re looking for correct example sentences to learn the usage of a word or when you’re trying to check if some phrase is correct, because you cannot trust the information you get from the general Web. In each case, you have to examine the source of the sentence to check if it’s trustworthy.
- Google routinely reports the wrong number of hits, especially for phrases. It may tell you that “I have a question for you” occurs on 1,600,000 pages, but the actual number is 473. This means you cannot trust the reported number of hits when you want to check if some phrase is correct. The only solution is to find the last search results page, but this can be hard if there are a lot of hits. (Bing used to be accurate, but now has the same issues.)
The Correct English search engine (based on Google) solves the first problem. It includes a subset of the Web — a hand-picked list of sources which are known to contain good English: online dictionaries, news sites, selected blogs and communities, Wikipedia, movie scripts, government sites, and others. Of course, the content is not 100% “pure”, but the quality is vastly better than on the general Web. Correct English contains practically no sentences written in bad English.
Yes, a proper corpus (such as the COCA) will give you much more searching power (for one thing, it lets you specify part-of-speech wildcards), but Correct English is quicker, easier to use and includes a much larger set of sentences. It does not show the number of hits, so you’ll have to click through to the last results page if you want to find out how common something is.
A definitive example of Google’s incorrect search result count (if Google’s own admission is not enough for you):
- eavesdropping “what did you hear” – “about 679,000 results”
- eavesdropping what did you hear – “about 111,000 results”
It is logically impossible for the first query to return more results than the second query. The second query should return more results because it includes pages which contain the words what, did, you and hear anywhere on the page, while the first query requires that they appear in that specific phrase.