The concept of an oracle for Turing machines assumes that the oracle answers Yes/No to a particular question $Q$, assuming that $Q$ is formulated as a bitstring on the oracle tape (instead of answering Yes/No, the oracle can generate the answer as a bitstring on the oracle tape).
Let $O$ denote an oracle that “encodes a well-order of order type $\omega_1^\text{CK}$” or, in other words, “has access to the well-order relation describing the well-ordering of $\omega_1^\text{CK}$ in terms of $\mathbb{N}$”. Can anyone explain what the set of questions for $O$ is and how a particular $Q$ (from the set of all possible questions) is converted to a bitstring?
You're right that we can't really use a structure as an oracle, per se; instead, we look at copies of the structure, that is, ways of representing the structure as a set of natural numbers. Incidentally, everything I say below applies to general countable structures.
A relation $E\subseteq\omega\times\omega$ is a copy of $\omega_1^{CK}$ if it defines a structure isomorphic to $\omega_1^{CK}$; that is, if $(\omega;E)\cong\omega_1^{CK}$. EDIT: If you prefer, you can think of a copy as a pair $(A; E)$ with $A\subseteq\omega$ and $E\subseteq\omega^2$ which is isomorphic to $\omega_1^{CK}$; for our purposes, this won't make a difference - see the comments below. It might be easier to think about this in a simpler case: consider the relation given by $mEn$ iff
$m$ is odd and $n$ is even, or
$m,n$ have the same parity and $m<n$.
This is a copy of the ordinal $\omega+\omega$: it's describing the linear order $$1\prec 3\prec 5\prec 7\prec 9\prec ...\prec 0\prec 2\prec 4\prec 6\prec 8\prec ...$$
Copies can be used as oracles in the obvious way, since a copy is just a set of pairs of natural numbers. Of course, there is no unique copy of $\omega_1^{CK}$, although we may argue that one copy or another is particularly "natural." The types of questions we ask in this context range across all copies of a structure; for example, the statement "$\omega_1^{CK}$ is not computable" is shorthand for "there is no computable copy of $\omega_1^{CK}$."