The ghost Gutzwiller ansatz (gGut) technique was shown to achieve accuracy comparable to dynamical mean-field theory at a much lower computational cost. However, gGut is limited by the bottleneck of computing the single-particle reduced density matrix. We develop a hybrid quantum-classical gGut technique that computes ground state properties of embedding Hamiltonians on a quantum computer using the quantum-selected configuration interaction (QSCI) algorithm. We find that the ground states of interest become sufficiently sparse as the number of ghost orbitals increases. We investigate QSCI’s performance using local unitary cluster Jastrow (LUCJ) ansatz with circuit cutting on IQM’s quantum hardware for up to 24 qubits. Our converged gGut calculations correctly capture the metal-to-insulator phase transition in the Fermi-Hubbard model. This was achieved using quantum samples to build a basis with as little as $1\%$ of the total CI states.