Ambient‐noise tomography (ANT) has become a well‐established method to image the crust and uppermost mantle structures in the past 15 yr. Having a good estimate of uncertainties of phase velocity dispersion measurements in ANT is critical as they can guide the level of data fitting in tomography. However, to date, there are still no systemic studies to evaluate these uncertainties. In this study, we obtain cross correlations with different stacking durations from 17 yr of ambient‐noise data recorded at 120 stations in the United States. We analyze the variations of signal‐to‐noise ratio (SNR) and phase velocities of cross correlations. We find that the uncertainties of phase velocities are affected by SNRs, interstation distances, and stacking durations. However, none of those three variables can be solely used as a proxy to estimate the uncertainties of phase velocity measurements. Based on our analysis, we graphically present empirical relations of uncertainties of phase velocity measurements as a function of SNR, interstation distance, and stacking duration. These relations can be employed as a guide to estimate phase velocity uncertainties in applications of ANT, assisting in evaluating the reliability of resulting models from ANT.