In basic quantum communication protocols one party creates quantum states and uses a quantum channel to transmit it to another party that performs immediately some measurement on it. This means, we effectively create correlated (classical) data between distant parties. In order to use the power of quantum mechanics, these correlation must show effects of quantum mechanics.
In the specific example of quantum key distribution one uses the correlations to distill a secret key in (classical) public discussion protocols e.g. via sifting, error correction and privacy amplification. We give a necessary condition for the success of any public discussion protocol: the observed correlations should allow to prove the presence of an internal, virtual state of entanglement in the distribution. This poses a first test whether any presented real quantum key distribution is indeed useful for the desired purpose. Moreover, a gap between the parameter regime of proven security of given realistic schemes and the regime of proven presence of vitual entanglement furthers the search for the optimal public discussion protocol.