Purpose. To report the accuracy of colposcopically directed biopsy in an internet-based colposcopy quality assurance programme in northern Italy. Methods. A web application was made accessible on the website of the regional Administration. Fifty-nine colposcopists out of the registered 65 logged in, viewed a posted set of 50 digital colpophotographs, classified them for colposcopic impression and need for biopsy, and indicated the most appropriate site for biopsy with a left-button mouse click on the image. Results. Total biopsy failure rate, comprising both nonbiopsy and incorrect selection of biopsy site, was 0.20 in CIN1, 0.11 in CIN2, 0.09 in CIN3, and 0.02 in carcinoma. Errors in the selection of biopsy site were stable between 0.08 and 0.09 in the three grades of CIN while decreasing to 0.01 in carcinoma. In multivariate analysis, the risk of incorrect selection of biopsy site was 1.97 for CIN2, 2.52 for CIN3, and 0.29 for carcinoma versus CIN1. Conclusions. Although total biopsy failure rate decreased regularly with increasing severity of histological diagnosis, the rate of incorrect selection of biopsy site was stable up to CIN3. In multivariate analysis, CIN2 and CIN3 had an independently increased risk of incorrect selection of biopsy site.