I doubt you'd be able to prove that it's rational, regardless of how it's constructed, since wouldn't that imply that it's also (for example) a real number?
Still, I don't see how any such definition could even be called a "definition of √-1" in the first place, since the resulting object wouldn't actually have the properties of √-1.
i tends to pop out as a construction so I suspect you can probably prove it is both rational and irrational or neither, depending on the fine print.
Here's a discussion: https://math.stackexchange.com/questions/823970/is-i-irratio...