"I have a question for you" Vs "I have a question to you"

What is the proper way of starting an email:

Hi X,

I have a question for you.

Or

I have a question to you.

Google fings 122M of the first version and 45M of the second one.Is the second one always incorrect?

Solution 1:

The idiom is overwhelmingly a question for (somebody). You'll sometimes meet the words a question to (someone), but usually in the context of put a question to (someone), where to is designating the indirect object the the verb put, not part of the noun phrase a question.

ie.

I have [a question for you].

but

I will put [a question] to you.

I guess that the majority of the 45M examples you found were from people who did not have English as their first language.

Solution 2:

SUPPLEMENTARY:
This is a methodological note raised by the form of your question, not an answer to the question itself (which has been quite adequately answered already).

• Never trust those numbers of hits Google reports. They vary widely, depending on where you post your search from (they told me 3,030,000 hits on "I have a question to you"!), and they are "estimates" which bear no relationship to the number of actual hits they can show you. I have on occasion been told Google had tens of thousands of hits on a particular phrase, which when followed up turned out to be ten pages or fewer: less than a hundred.

• A more reliable method of assessing usage is through Google Ngrams. This only surveys printed sources, so it doesn't catch much colloquial use (except the heavily stylized dialogue in fiction and scripts); but the graphs it provides are tolerably reliable. Here's an informative Ngram for your two phrases, "have a question for you" and "have a question to you":

  

  Note the box: "not found: have a question to you", confirming the answers you have been given.

  You do have to be careful with your search terms here (as you do with Google, too, of course). You can't for instance rely on a comparison of "question for you/question to you", because this will pick up a lot of irrelevant hits on "put a question to you"

  Also note that the scale is a percentage. A large spike can be caused by a single book in a relatively small number of samples. And that single book might be the result of an OCR error. Results from 1900 onwards are less likely to feature recognition errors, and results from recent years will feature a larger number of books in the sample.