Background The dramatic clinical consequences of anastomotic leakage in gastrointestinal surgery can be reduced by a diverting stoma or drainage of the peri-anastomotic area. Currently, the surgeons' clinical judgement is of major importance in decision making, but reliable data of the diagnostic accuracy are lacking. In this prospective clinical study, the surgeons' predictive accuracy for anastomotic leakage was evaluated. Materials and methods In 191 patients undergoing colorectal resection with anastomosis, the risk for anastomotic leakage was determined by the surgeon on the basis of a visual analogue scale (VAS). This risk assessment was compared to the actual occurrence of anastomotic leakage post-operatively. Results A total of 26 (13.6%) patients showed anastomotic leakage. The surgeons' median predicted leakage rate was 7.1% in anastomoses >15 cm from the anal verge and 9.5% ≤15 cm (sensitivity 38/62%, specificity 46/52%). Diagnostic accuracy was not influenced by the surgeons' training level (VAS score, surgeons 7.8% vs assistant surgeons 8.5%, p=0.96, sensitivity 41% vs 44%, specificity 59% vs 48%, p=0.20). Conclusion The surgeons' clinical risk assessment appeared to have a low predictive value for anastomotic leakage in gastrointestinal surgery. The low a priori risk of anastomotic leakage of 14% resulted in a low post-test odds (11%) of correct prediction of anastomotic leakage. This warrants the ongoing search for a better diagnostic test of anastomotic leakage to prevent morbidity and mortality.