Belnap and Steel (1976, 5) define a question as presupposing a statement if and only if the truth of the statement is a logically necessary condition for there being a true (i.e., correct) answer to the question. ... To deny a presupposition of a question is to give a corrective answer to the question, but most theorists join Belnap and Steel in not counting corrective answers as direct answers.
I suppose what they have in mind is the pragmatic context of, say, "Have you stopped milking your giraffes?" If you don't have any giraffes, or if you never even started milking them, you commit a pragmatic gaffe if you say either, "Yes," or, "No," or even, "Maybe," here.
But that's just a pragmatic issue; more strictly, if you have no giraffes/never milked them to begin with, then you'd be "correct" to say, "No," since it's "logically" true that you can't stop doing something unless you've started doing it beforehand. I would expect the presupposition-relation to hold faster than just to our pragmatic expectations as such.
Now, there are also questions like, "If the last aardvark took over the universe, would every word starting with one a be rewritten with two a's in that position, e.g. America would become Aamerica, etc.?" Now a question having a pseudo-assertoric antecedent seems like "involving a presupposition," and at least for a sort of indexical reason, the erotetic consequent could appear to be deficient in intelligibility without us referring back to the meaning of the antecedent. E.g., if someone just asked, "Would every word starting with one a...?" we would wonder what the conditions of "Would" there, are.
Are conditional questions the only ones involving "semantic," not pragmatic, presuppositions; or are there some questions involving no presuppositions whatsoever, e.g. imperative questions like, "Count to 10?" Or might one say of even, "Count to 10?" that it "presupposes" the existence of counting and the number 10?