The ultraviolet (UV) continuum slope β, typically observed at z 7 in Hubble Space Telescope (HST) WFC3/IR bands via the J − H colour, is a useful indicator of the age, metallicity and dust content of high-redshift stellar populations. Recent studies have shown that the redward evolution of β with cosmic time from redshift 7 to 4 can be largely explained by a buildup of dust. However, initial claims that faint z 7 galaxies in the Hubble Ultra Deep Field WFC3/IR imaging (HUDF09) were blue enough to require stellar populations of zero reddening, low metallicity and young ages, hitherto unseen in star-forming galaxies, have since been refuted and revised. Here we revisit the question of how best to measure the UV slope of z 7 galaxies through source recovery simulations, within the context of present and future ultra-deep imaging from HST. We consider how source detection, selection and colour measurement have each biased the measurement of β in previous studies. After finding a robust method for measuring β in the simulations (via a power-law fit to all the available photometry), we remeasure the UV slopes of a sample of previously published low-luminosity z 7 galaxy candidates. The mean UV slope of faint galaxies in this sample appears consistent with an intrinsic distribution of normal star-forming galaxies with β −2, although properly decoding the underlying distribution will require further imaging from the ongoing HUDF12 programme. We therefore go on to consider strategies for obtaining better constraints on the underlying distribution of UV slopes at z 7 from these new data, which will benefit particularly from the addition of imaging in a second J-band filter: F140W. We find that a precise and unbiased measurement of β should then be possible.