We prove that maximal probability of cheating in quantum signing protocol of Paunković, Bouda and Matheus behaves as $O\big(1/\sqrt{N}\,\big)$ for large number of qubits $N$. This confirms a conjecture that was based on numerical evidence.