Objective This study compares the accuracy and reliability of WebCeph (web-based program for cephalometric analysis) with the AutoCAD computer software. Materials and methods A sample of pretreatment digital lateral cephalograms of 50 orthodontic patients was analysed with WebCeph and AutoCAD software (as a standard measure). On each cephalogram, 17 landmarks and 11 measurements were marked and performed as skeletal, dental, and soft–tissue parameters. We used six angular and five linear measurements. A paired t-test was used to assess the systematic bias. The intraclass correlation coefficient (ICC) and Bland–Altman plot with linear regression analysis were used to assess the agreement between the two methods. Results There was adequate reproducibility for the measurements with both WebCeph and AutoCAD. The paired t-test showed statistically significant differences for five angular and two linear measurements (P < 0.05). The ICC test between WebCeph and AutoCAD revealed very good to excellent agreement for all measurements, except for the lower incisor to mandibular plane angle. The Bland–Altman plot visually showed a relatively acceptable limit of agreement for three angular and two linear measurements only, and the linear regression analysis revealed a significant proportional bias between the two methods for four angles and the upper lip-Esthetic line (U Lip-E Line). The systematic bias and level of agreement improved with the use of the semi-automatic WebCeph. Conclusions Different problems, such as poor landmark identification/soft tissue tracing and inconsistency of measurements, are inherent to the automatic WebCeph. The semi-automatic WebCeph can overcome some limitations of the automatic WebCeph; however, it should be used for cephalometric analysis with a great deal of caution.