The next generation of wide-field spectroscopic redshift surveys will map the large-scale galaxy distribution in the redshift range 0.7 <= z <= 2 to measure baryonic acoustic oscillations (BAO). The primary optical signature used in this redshift range comes from the [OII] emission line doublet, which provides a unique redshift identification that can minimize confusion with other single emission lines. To derive the required spectrograph resolution for these redshift surveys, we simulate observations of the [OII] (lambda lambda 3727, 3729) doublet for various instrument resolutions, and line velocities. We foresee two strategies for the choice of the resolution for future spectrographs for BAO surveys. For bright [OII] emitter surveys ([OII] flux similar to 30 x 10(-17) erg cm(-2) s(-1) like SDSS-IV/eBOSS), a resolution of R similar to 3300 allows the separation of 90 percent of the doublets. The impact of the sky lines on the completeness in redshift is less than 6 percent. For faint [OII] emitter surveys ([OII] flux similar to 10 x 10(-17) erg cm(-2) s(-1) like DESi), the detection improves continuously with resolution, so we recommend the highest possible resolution, the limit being given by the number of pixels (4k by 4k) on the detector and the number of spectroscopic channels (2 or 3).