We present a thorough investigation of stellar photometry based on HST imaging of crowded fields at 85 and 10 arcsec from the centre of the high-surface brightness elliptical M32. The Principal Investigators of the present archive data have elsewhere presented an impressive colour-magnitude diagram of the field at 85 arcsec. Based on the same data we enlarge on their photometric analysis and supplement with error estimators that more clearly show the implications of severe image crowding on the stellar photometry. We show that the faintest stars (I>25.0, V>26.0) are found too bright by several tens of a magnitude. For the field at 10 arcsec we conclude that it is not possible to obtain reliable stellar photometry, standard deviations being larger than 0.4 mag. Artificial-star experiments show that only very few of the brightest stars of the luminosity function can be expected to represent single objects, the majority being either spurious or not as bright as measured.