Objective To compare the intercycle variation of serum anti‐Mullerian hormone (AMH) and antral follicle count (AFC) measurements over four consecutive menstrual cycles. Design Observational study with secondary analysis using data from two previous randomized controlled trials. Patients Seventy‐eight women from two previous randomized trials on the effect of dehydroepiandrosterone pretreatment on ovarian response in women undergoing in vitro fertilization (IVF) treatment. Measurements The intraclass correlation coefficients (ICC) for AFC and AMH across the four study cycles, as well as their predictive performance on poor ovarian response, were compared. Results No significant difference was observed in AMH (p = .608) across the four study cycles. AFC was significantly higher at 4 weeks before ovarian stimulation compared with 0, 8 and 12 weeks before ovarian stimulation (p < .05, Conover posthoc test). Both single‐measures and average‐measures ICC were significantly higher with AMH than with AFC. The areas under the receiver operating characteristic curve of the four AFC measurements in predicting poor ovarian response (defined as three or less oocytes retrieved) in the IVF cycle ranged from 0.657 to 0.743 with no significant difference (p > .05) among the four cycles, whereas those of the four AMH measurement ranged from 0.730 to 0.780 with no significant difference (p > .05) among the four cycles. Conclusions Although both AFC and AMH are good predictors of ovarian response, intercycle repeatability was significantly better with serum AMH than AFC measurement. Both have no significant difference in their predictive performance on poor ovarian response when assessed within three months before IVF treatment, hence allowing pre‐IVF assessment at more flexible timing.