We present the results of a search for dwarf novae (DNe) in globular clusters (GCs). It is based on the largest available homogeneous sample of observations, in terms of the time‐span, number of observations and number of clusters. It includes 16 Galactic GCs and yielded two new certain DNe: M55‐CV1 and M22‐CV2. All previously known systems located in our fields were recovered, too. We surveyed M4, M5, M10, M12, M22, M30, M55, NGC 288, NGC 362, NGC 2808, NGC 3201, NGC 4372, NGC 6362, NGC 6752, ω Centauri (NGC 5139) and 47 Tucanae (NGC 104). The discovery of two DNe, namely M55‐CV1 and M22‐CV2, was already reported by Kaluzny et al. and Pietrukowicz et al., respectively. In the remaining 14 GCs, we found no certain new DNe. Our result raises the total number of known DNe in the Galactic GCs to 12 DNe, distributed among seven clusters. Our survey recovered all three already known erupting cataclysmic variables (CVs) located in our fields, namely M5‐V101, M22‐CV1, and V4 in the foreground of M30. To assess the efficiency of the survey, we analysed images with inserted artificial stars mimicking outbursts of the prototype DNe SS Cygni and U Geminorum. Depending on the conditions, we recovered between 16–100 per cent of these artificial stars. The efficiency seems to be predominantly affected by duty cycle/time‐sampling and much less by distance/magnitude. Except for saturated tiny collapsed cores of M30, NGC 362 and NGC 6752 (and also the dense core of NGC 2808), crowding effects in the V band were avoided by our image subtraction technique augmented with auxiliary unsaturated B‐band images. Our results clearly demonstrate that in GCs common types of DNe are very rare indeed. However, great care must be taken before these conclusions can be extended to the CV population in GCs.