Luteal-phase ovarian stimulation has been proved to be feasible for producing competent oocytes/embryos and achieving live births, yet there is no standardized stimulation protocol for luteal-phase ovarian stimulation (LPS). The aim of this study was to explore the optimal timing of gonadotropin initiation in the LPS protocol for poor ovarian responders. This was a retrospective cohort study conducted in the reproductive medicine center of a tertiary hospital. A total of 327 poor responders fulfilling Bologna criteria underwent LPS with IVF/ICSI treatment. HMG and letrozole were administrated after ovulation. Patients were stratified into three groups according to the gonadotropin start day: early, early-mid, and mid-late luteal phase. A freeze-all strategy was performed for all cycles. The duration of ovarian stimulation, total gonadotropin dose, number of oocytes retrieved, implantation rate, clinical pregnancy rate, and live birth rate after frozen/thawed embryo transfer cycles were included for evaluation. The group accepted ovarian stimulation in the earlier phase tended to have a shorter duration of ovarian stimulation [8 (7,10) in early luteal group, 9 (8,10.25) in early-mid luteal group, and 11 (10,12) in mid-late luteal group; P <0.001] and lower gonadotropin consumption [1993.35 ± 720.31, 2282.73 ± 703.38, and 2764.83 ± 722.26, respectively; P <0.001]. Logistic regression and multiple linear regression were used to assess the associations between the phase of gonadotropin initiation and duration of ovarian stimulation (or total gonadotropin dose) by adjusting for confounding factors. Compared with the early luteal group, longer ovarian stimulation(>9 days) was more likely to occur in the early-mid and mid-late luteal groups, with the adjusted odds ratios 0.584 (0.327-1.042) and 0.116 (0.049-0.271), respectively (P-trend<0.001). Delayed gonadotropin initiation showed an 113.200 IU increase (95%CI: 70.469, 155.930) per-day in the total gonadotropin dosage. Meanwhile, there were no significant differences in the mean number of oocytes, utilizable embryos, pregnancy outcomes among three groups. Although the timing of gonadotropin initiation is not associated with pregnancy outcomes, earlier initiation of gonadotropin therapy after ovulation was associated with a shorter duration of ovarian stimulation and lower gonadotropin consumption in poor responders in LPS.