Abstract Objective To compare the efficacy of the long GnRH agonist and the fixed GnRH antagonist protocols in IVF poor responders. Study design This was a randomized controlled trial performed in the Iakentro IVF centre, Thessaloniki, from January 2007 to December 2011, concerning women characterised as poor responders after having 0–4 oocytes retrieved at a previous IVF cycle. They were assigned at random, using sealed envelopes, to either a long GnRH agonist protocol (group I) or a GnRH antagonist protocol (group II). Results Overall 364 women fulfilled the inclusion criteria and were allocated to the two groups: finally 330 participated in our trial. Of these, 162 were treated with the long GnRH agonist protocol (group I), and 168 with the fixed GnRH antagonist protocol (group II). Numbers of embryos transferred and implantation rates were similar between the two groups (P=NS). The overall cancellation rate was higher in the antagonist group compared to the agonist group, but the difference was not significant (22.15% vs. 15.2%, P=NS). Although clinical pregnancy rates per transfer cycle were not different between the two groups (42.3% vs. 33.1%, P=NS), the clinical pregnancy rate per cycle initiated was significantly higher in the agonist compared to the antagonist group (35.8% vs. 25.6%, P=0.03). Conclusions Although long GnRH agonist and fixed GnRH antagonist protocols seem to have comparable pregnancy rates per transfer in poor responders undergoing IVF, the higher cancellation rate observed in the antagonist group suggests the long GnRH agonist protocol as the first choice for ovarian stimulation in these patients.